Resources

Demand for agentic security that actually works in complex enterprise environments has never been higher, and today we're excited to take a meaningful step forward in meeting it

We're excited to announce that Legion Security has partnered with Optiv to become an Authorized Partner to help enterprises stop talking about the same-old-problem, and start putting AI to work. Security teams are under pressure that doesn't need a lot of explaining. Analysts, engineers, and practitioners are being asked to do more with less; more alerts, more tools, more threat surface, and fewer people to manage it all. AI was supposed to be the great equalizer, and the promise of the AI SOC was compelling: automate the noise, free up your people, let machines handle the volume.

The reality has been… more complicated.

Most AI security tools were built generically for a generic security team in a generic enterprise. One problem with this is… what is an average security team? Every large organization has processes that are entirely their own: workflows built around a specific stack, custom tools that were built and tuned over long stretches, tribal knowledge accumulated over years, investigation procedures tuned to their environment, their risk tolerance, their regulators, their customers.

Heavy API integrations try to stitch it together but end up slow, brittle, and context-poor (at best). And agents that operate inside a black box create exactly the kind of trust deficit that makes security leaders hesitate to hand anything off at all.

This is the gap Legion was built to close.

A Different Approach to Agentic Security in the Enterprise

The premise of Legion is straightforward: nobody knows your security operations like you do. Our platform doesn't arrive with assumptions about how your team should work. Instead, it observes and learns from how your team actually works; across your tools, your workflows, your most repetitive processes and your most bespoke ones, and then uses that knowledge to build optimized AI agents that operate within the context of your organization.

We don’t require integrations for full contextual awareness. We’re an open book (no black box) that leans on our browser-based approach to see what your analysts see and do, learns what they know, and earns YOUR trust before taking action.

The result is agentic security that can actually scale in the enterprise — not by replacing how teams work, but by amplifying it.

The Imperative for Partnering with Optiv

Becoming an Optiv Authorized Partner matters because of what Optiv represents to the enterprise security buyer. Optiv works with organizations that have mature, complex security programs; exactly the kind of environment where Legion's approach of learning from bespoke processes is most valuable.

Enterprise security leaders look to trusted advisors to help them evaluate fit, plan implementation, and optimize outcomes over time. Optiv's position in the market as an integrator with deep relationships and deep domain expertise makes them uniquely positioned to bring best-in-breed solutions to the organizations that need it most and to help them get maximum value from it.

This partnership reflects something we're hearing consistently in the market: enterprises want agentic security, but they want it on their terms. They want AI that understands their environment before it acts in it. They want partners who can help them think through where automation should start, how to build confidence in the system over time, and how to expand from their first use cases into a broader program.

That's exactly what this partnership is designed to deliver.

What It Signals More Broadly

The Optiv partnership is a data point in a larger trend. Channel partners; the integrators, MSSPs, and advisors who sit closest to enterprise security buyers, are increasingly being asked about agentic security. Their clients want to know what's real, what's ready, and what actually works in complex environments.

For Legion, this is an important milestone in building the ecosystem that enterprise agentic security requires. We're grateful to the Optiv team for their partnership and excited about what we'll build together. And for enterprise security leaders who have been watching the agentic security space and wondering what a path to trusted AI adoption actually looks like, we'd love to show you.

Interested in learning how Legion Security and Optiv can help your organization automate, scale, and elevate your security posture? Get in touch.

‍

I was there, I sat in every SOC seat out there…

A SOC analyst grinding through alert queues at 2am. Part of an Incident Response team leading running war rooms. A SOC manager in Monday morning stand-ups asking what we learned this week while staring at blank faces.

Every single role. Every single day. And the one thing that never changed across any of them?

The insights, recommendations, self improvement, the de-facto SOC continuous improvement action items were disappearing. Seating documented in a case log for no one to action upon, trapped inside closed tickets that live in a backlog nobody rarely reopens.

I know the why and I feel the overwhelming operations, which is  why I’m offering a practical solution for how to continuously improve your SOC with the valuable insights coming out of your investigations.

The Hidden Goldmine You're Sitting On

Every ticket your team closes tells a story. It's not just that an alert fired, then an analyst investigated and eventually closed. There are powerful signals buried in those notes, whether it's a tool with overly noisy alerts, a gap in your email gateway rules, or the same user clicking a phishing link for the third month in a row.

Your tier 1 all the way to your tier 5 analysts and IR responders are generating intelligence every single shift and with every single incident. They know things and they're writing them down. It's useful information but these notes get buried and never read again.

It's a sad truth... I know because I've been in those weekly SOC meetings, I was running them.

It's not a people problem, rather, it's a system problem.

The Weekly Report Trap

The thing people look to as the standard fix is the weekly report. In theory it's elegant: senior analysts summarize the week, extract the learnings, feed them back into tier 1 runbooks and detection improvements. On paper, it's a proper feedback loop.

In practice, it becomes the task that either gets rushed on Friday afternoon or simply doesn't happen. It's for good reason too! Your senior analysts are already stretched because on top of everything they need to do for their jobs, they're also being asked to synthesize everything in themes. You either get a half-hearted copy-paste of ticket titles, or, more likely, you get nothing.

Teams try rotation where everyone takes a turn on the ferris wheel. But in doing so, you face losing important insights and information, not to mention a lack of consistency.

Now add a follow-the-sun operation to this. APAC closes tickets while EMEA is asleep. EMEA handles incidents while Americas is offline. By the time anyone tries to compile a summary, they're working with fragments. Nobody has the full picture. The patterns that only emerge when you look across all shifts stay invisible.

Wait, Can't AI Can Solve This Pretty Easily? 

When capable LLMs became available, I thought this was finally solved. Just feed all the investigation summaries in, ask for a weekly report. Done? Not so fast... here's what actually happened.

First attempt: I gave the best LLM models that money can buy more than 250 investigation summaries and asked for a consolidated report. But what I got back was a mess.

What I saw were recommendations repeated five times just with slightly different wording. Severity assessments that made no sense and my “favorite” recommendations that are not feasible, for example “Tune your EDR machine learning to reduce false positives of macro xlsx files”.

No traceability whatsoever, no way to tie anything back to the original investigation and forget about cross referencing with similar recommendations.

Second attempt: I went deep on prompt engineering. Longer prompts. More detailed. With examples. The results improved marginally, but the ceiling was surprisingly low.

The fundamental issue is that when you dump a large context with complex requirements into a single LLM call, it can't hold everything in working memory. It forgets constraints from earlier in the prompt. It hallucinates connections between unrelated incidents. Severity levels come out inconsistent.

One-shot approaches get you mediocre fast. They don't get you useful.

The Breakthrough: Think Multi-Step, Not Prompt

The shift that changed everything was stopping thinking about this as one task and starting to think about it as a multi-step pipeline.

When an experienced analyst writes a weekly report, they don't try to do it all at once. They read, they group, they prioritize, they write. Multiple steps. Each one is different.

So I built it that way.

The 6-step pipeline

Step 1: Classification

The first step does one thing and one thing only. It extracts and categorizes recommendations from raw investigation summaries. It looks for whatever your analysts call them: Recommendations, Do Better, Action Items, Next Steps. It pulls each one out and assigns it to a category: detection, prevention and process improvements.

No dedupe. No severity. Just extraction, done well.

Step 2: Feasibility Assessment

Now we evaluate each recommendation against practical reality. Can this actually be implemented? Is it a quick win or a multi-quarter project? Does it require resources you don't have?

This is also where web search earns its keep. When a recommendation references a specific product or vendor, the model can look up current best practices, product documentations, tech community discussions and verify the suggested configuration actually exists and is supported. Without this, you get generic, often infeasible advice. With it, you get grounded recommendations.
Make sure to use an LLM model that has web search capability via API calls.

Step 3: Citation Attachment

Before touching deduplication, every recommendation gets linked back to its source investigation. This is non-negotiable for a report anyone will actually act on. When a SOC manager reads and SOC teams attempt recommendation implementation, they need to know which investigations triggered that and value with volume justification to it. Otherwise it's just noise or worse, it might break business operations.

Step 4: Deduplication

Three analysts working three separate investigations but same use case, all recommend the same prevention improvement. Without deduplication, you get three entries saying the same thing with slightly different wording. With it, you get one consolidated recommendation that shows it came from three independent investigations, which is actually a stronger signal.

Citations from all source recommendations get merged. Nothing is lost.

Step 5: Severity Classification

Now, with duplicates consolidated, we can assign severity levels that actually mean something. The model evaluates security impact per your instructions, weights and SOC defined severities for each use case. Not how urgent did the analyst feel when writing this, but what is the actual risk if this doesn't get addressed built on your SOC knowledge base.

Separating this from extraction forces objectivity. If you try to assign severity while also pulling recommendations from raw notes, the analyst's tone bleeds in and skews the assessment.

Step 6: Report Generation

Everything feeds into the final structure. The model has category breakdown, feasibility assessments, severity levels, citation references. It produces a coherent report with an executive summary and recommendations sorted by severity, with enough context to actually act on. Also comparing recommendations week on week to get remediation/implementation progress for repeated action items.

Add another layer of disregard recommendations and you have a magnificent mechanism.

No LLM at this stage, actually. It's programmatic and deterministic. It assigns citation letters for easy grounding and reference of recommendation with feasibility (A, B, C...), builds the reasoning section for each recommendation, and outputs clean JSON ready for whatever you want to do with it.

Why This Architecture Actually Works

The goal is to achieve focused context at each step. Instead of one massive prompt juggling ten objectives, each step gets only what it needs. Fewer constraints to forget.

Modular iteration is the name of the game here. When severity ratings were inconsistent, I refined only the severity prompt. When analysts switched from Recommendations to Do Better as their section header, I updated only the classification step and nothing else broke.

Inspectable intermediate outputs. Between every step, results are saved. If something looks wrong in the final report, you can trace back through the pipeline and find exactly where it broke. Debugging is possible, which is not nothing.

Web search in the right place. Not as a general capability, but specifically in the feasibility step where it does the most work. Validating that a recommended configuration actually exists changes the quality of the output completely.

The Payoff

Your analysts don't change anything, they can run the same investigations, keep the same ticket notes they're already writing. The pipeline simply runs against their existing documentation.

The output is consistent. Same structure, same categories, same severity criteria, every week. You can compare week over week and actually spot trends. You can see if the same recommendations keep surfacing, which means they're not getting actioned, which is itself a signal.

The feedback loop that should have existed closes automatically. Tier 2 findings reach tier 1. Detection gaps surface. The Monday morning question about what we learned has an answer.

Build it or use it

Building this right takes time. Getting prompts tuned for the variety in how analysts write, handling edge cases, making it robust across different ticketing systems. It's not weekend work.

If you want to build it yourself: start with extraction only. Get that reliable first. Then add deduplication. Then severity. Don't try to build the whole thing at once.

If you'd rather not build tooling while also running a SOC, this is exactly what we built at Legion Security. Already tuned across real SOC environments, connected to your existing ticketing system, your analysts change nothing.

Either way: stop burying the intelligence your team generates every day.

Your team is learning constantly. Those lessons deserve to surface.

Written by someone who's been the analyst, the IR lead, and the manager staring at the empty Monday morning whiteboard.

‍

Legion Security is Now Available on Google Cloud Marketplace

Security operations were built around human investigators. Skilled analysts, working manually across dozens of tools, piecing together evidence, making judgment calls, closing cases. But as alert volumes outpaced human capacity, institutional knowledge became a bottleneck, and the complexity of the modern enterprise made scaling impossible. The industry responded with more headcount, more tools, more automation. None of it solved the fundamental problem.

Legion introduces a different operating model entirely. 

What Legion Does

Legion observes how your analysts operate when running real investigations, learning your organizational context, tools, past cases, playbooks, runbooks and all other tribal knowledge in order to understand what an optimal investigation looks like for your environment. This is then turned into an easily editable and audible workflow which can be automated when you’re ready. Powered by Google Cloud's Gemini models, each workflow is executed by AI agents that reason through the evidence and provide a verdict and even remediate. This is all accomplished with no manual playbook writing or need to document predefined rules.

But legion goes well beyond workflow creation. As Legion builds trust in its performance, teams can choose to keep a human in the loop to approve every decision or have Legion operate fully autonomously reducing MTTR eliminating MTTA, allowing analysts to focus on more novel investigations that are becoming more and more common in the world of AI. 

Memory: The Compounding Advantage

Every investigation Legion conducts makes it smarter. A persistent memory layer continuously captures context from previous cases, your SOC knowledge base, and direct analyst feedback, feeding all of it back into future investigations and decisions. Institutional knowledge that once lived in the heads of your most experienced analysts becomes a permanent, improving organizational asset. The more Legion works, the better it gets. That's not a feature. That's a compounding strategic advantage.

Zero Integrations. Immediate Value.

Most security automation platforms fail at the same hurdle: integrations. Enterprises face months of API work, custom connectors, and professional services before anything runs in production, or are forced to adopt entirely new tools and processes, something most complex enterprises simply can't do.

Legion operates natively in the browser, which means it works across your entire security stack, from threat intel platforms to legacy internal tools, without any API configuration. If your analysts can open it in a browser, Legion can learn from it, generate workflows from it, and execute investigations through it.

Proven Results at Scale

The impact Legion delivers isn't theoretical:

As the head of Security at Virgin Money put it, Legion is “like evolving from handcrafted systems to precision manufacturing aligned to our flow (except) faster, repeatable and secure”.

Legion works with the worlds largest enterprises and delivers strong results: 

• A large insurance organization automated 24,000 investigations and cut mean time to respond from 20 minutes to 2 minutes.
• WELL Health Technologies reduced investigation times by 81%, allowing existing analysts to handle significantly higher alert volumes without additional headcount.
• The University of Tulsa cut investigation times in half, enabling their team to overcome capacity limits with the staff they already had.

Across deployments, Legion reduces mean time to investigate by up to 85% and response times by up to 90%.

Built on Google Cloud

Legion's integration with Google Cloud goes deeper than the Marketplace listing. The platform runs on Google Cloud infrastructure and leverages Gemini models to power its AI reasoning, combining Legion's browser-native architecture with Google Cloud's security, scale, and model quality.

For organizations already invested in Google Cloud and Google SecOps, Legion extends that ecosystem directly into the analyst workflow.

Who It's For

Legion is purpose-built for enterprise security operations teams, CISOs, VPs of Information Security, SOC Directors, and Security Operations Managers at organizations running in-house SOCs. If your team is dealing with any of the following, Legion was built for you:

• Alert volumes that have outpaced your team's capacity
• Analyst burnout from manual, repetitive investigation work
• Institutional knowledge that walks out the door when senior analysts do
• Automation gaps caused by complex integration requirements

Available Now on Google Cloud Marketplace

Legion Security is available today on Google Cloud Marketplace, allowing customers to apply their spend toward their annual Google contract and simplify procurement. For security teams ready to move beyond the limits of traditional operations, this is where that transformation begins.

‍

Introduction

When automating modern security investigations with LLM-based agents, to conduct multi-step security investigations, a typical workflow begins with an alert - say, a reported phishing email - and the agent iteratively queries tools such as Microsoft Defender Threat Explorer, Splunk, or CrowdStrike to gather evidence, assess scope, and recommend containment actions.

At each step, the agent receives query results containing raw IOCs: sender addresses, embedded URLs, source IPs, recipient domains, and device hostnames. It must reason about these indicators, decide whether to refine its search or conclude the investigation, and return its findings as structured output.

This works well for short investigations. But as the number of steps grows, an issue emerges: the agent's context window fills with repeated, verbose indicator values, the model begins echoing raw IOCs inconsistently, and the structured outputs it produces become increasingly fragile.

[Figure 1: High-level architecture of an AI-driven investigation agent.]

Consider a phishing investigation that proceeds through four steps:

1. Initial query: Search for emails from a reported sender to a specific recipient. The results contain the sender's email address and a handful of URLs.
2. Scope expansion: Search for all emails from the same sender across the organization. The results return 22 emails with SharePoint URLs, tracking links, and font-file references.
3. URL analysis: Search by specific URLs found in step 2. Additional domains and redirects surface.
4. Conclusion: The agent summarizes its findings and lists all relevant IOCs.

By step 4, the agent's prompt contains the full history of steps 1 through 3 - including every raw URL, email address, and domain mentioned in each step's results and the agent's own reasoning. Some of these URLs are long tracking links with base64-encoded parameters, easily exceeding 200 characters each.

This accumulation created three concrete problems.

1. Token bloat. Raw IOC values, particularly URLs with tracking parameters and encoded payloads, consumed a disproportionate share of the context window. A single newsletter email might contain 30+ URLs, each repeated in the query results, the agent's reasoning, and the indicators list, tripling the token cost per IOC, per step.
2. Over-reporting. When asked to list relevant indicators, the model would frequently dump every IOC it had ever seen into the response - even when the current step involved only one or two. In one case, an agent listed all 145 email addresses from its registry when the current query concerned a single sender.
3. Structural fragility. Query results from security tools sometimes contained comma-separated URL lists embedded in strings. When the model attempted to reproduce these in its JSON output, it produced malformed structures - unescaped commas, broken string boundaries, and invalid nesting. In our baseline evaluation, only approximately 80% of model responses parsed as valid JSON.

Our Approach

We addressed these problems with a three-part system:

1. A unified IOC manager that extracts and indexes indicators.
2. An IOC prompt adjustment that instructs the model on how to use indexed references.
3. A preprocessing step that cleans malformed tool output before it reaches the model.

IOC Extraction and Indexing

The core of the system is an IOC manager that maintains a registry of all indicators encountered during an investigation. When new text enters the pipeline - whether from tool query results or from the agent's own prior reasoning - the manager scans it using a set of type-specific patterns covering URLs, email addresses, IPv4 addresses, file hashes, hostnames, and domains.

Each newly discovered IOC is assigned a compact symbolic reference following a consistent naming convention: the first email becomes EMAIL01, the first URL becomes URL01, the second domain becomes DOMAIN02, and so on. The original value is stored in the registry, and all occurrences in the text are replaced with the corresponding reference.

Extraction order matters. URLs are processed first, because a URL contains both a domain and potentially an IP address. By extracting URLs before domains and IPs, we prevent the system from fragmenting a single indicator into multiple overlapping entries. 

The manager also performs selective extraction. In a typical prompt, the first section contains static task instructions - tool descriptions, output format specifications, and investigation guidelines. IOC extraction is applied only to the dynamic sections (step history and query results), leaving instruction text unchanged. This prevents false positives from example IOCs embedded in the prompt template.

Deduplication is handled through a value-to-reference mapping. If the same IOC appears in step 1 and again in step 3, it receives the same reference both times, ensuring consistent tracking across the entire investigation.

[Figure 3: The IOC extraction pipeline.]

IOC Prompt Adjustment

Extraction alone is not sufficient. Even when the input prompt uses symbolic references, the model may revert to generating raw IOC values in its output, particularly if the system prompt or prior conversation history contains raw values, or if the model has seen the actual value during context processing.

To address this, we developed an IOC prompt adjustment, a compact structured appendix appended to the user prompt that explicitly instructs the model on how to handle IOCs. The adjustment establishes three rules:

1. In reasoning: Always use symbolic references. Never write raw email addresses, URLs, IP addresses, or domains. Instead of writing a raw sender address followed by a description of the campaign, use the corresponding reference identifier throughout.
2. In the indicators field: Distinguish between known and new IOCs. For indicators already present in the registry, use the symbolic reference. For indicators being reported for the first time, newly discovered in the current step's results, use the actual value, so it can be added to the registry for subsequent steps.
3. Relevance filtering: Only include IOCs that are directly relevant to the current investigation step. Do not copy all registry entries into every response.

The IOC prompt adjustment includes a populated copy of the current IOC registry, mapping each reference to its actual value, so the model can look up identifiers when constructing its reasoning. It also provides correct and incorrect examples, a validation checklist, and explicit rejection criteria.

We tested two versions of the IOC prompt adjustment: a comprehensive version with extensive examples and redundant emphasis, and an optimized version that distills the same rules more concisely. Both achieved equivalent compliance rates, suggesting that clarity of instruction matters more than volume of repetition.

‍

Input Preprocessing

The second component addresses a problem upstream of the model: malformed tool output. Security tool APIs sometimes return URL lists as comma-separated values within a single string field, rather than as properly structured arrays. When passed through to the model as-is, these malformed strings caused structured output generation failures.

Our preprocessing step detects comma-separated URL patterns in query results and reformats them into clean, numbered lists before the text reaches the model. This small transformation, applied before IOC extraction, resolved the structured output validity issue independently of the other components.

Evaluation

Setup

We evaluated the system using 10 real-world investigation traces captured from production. Each trace represents a complete phishing investigation conducted through several security tools, containing the system prompt, user prompt with step history, and the raw query results that the model must reason about.

For each trace, we ran 10 iterations with the same prompt configuration, measuring two metrics:

• JSON validity: Whether the model's response parsed as valid a structured output.
• IOC reference compliance: Whether the response used symbolic references exclusively in its reasoning field (no raw IOC values) and correctly distinguished between known references and new actual values in its indicators field.

We tested four configurations to isolate the contribution of each component.

‍

Results

JSON validity. The baseline configuration produced valid JSON in approximately 80% of responses. Adding URL preprocessing alone brought this to nearly 100%, confirming that malformed tool output - not model capability - was the root cause of parsing failures. All configurations that included the enforcer achieved 100% validity.

IOC compliance. Without the IOC prompt adjustment, the model never spontaneously adopted symbolic references, compliance was 0% regardless of whether the input text had been processed by the IOC manager. With the prompt adjustment, compliance jumped to 100% across all traces and iterations. This held for both the comprehensive and optimized prompt adjustment variants.

Component independence. The results reveal a clean separation of concerns: URL preprocessing fixes JSON validity, the IOC manager fixes IOC compliance, and provides the underlying registry and extraction infrastructure that makes both possible.

Qualitative Observations

Beyond the quantitative metrics, we observed several qualitative improvements:

• Reduced prompt size. Replacing verbose URLs (some exceeding 200 characters) with compact references, meaningfully reduced token consumption in the step history, particularly for investigations involving newsletter or marketing emails with numerous tracking links.
• Consistent cross-step tracking. The registry ensured that the same IOC received the same reference throughout the investigation, this is particularly helpful with IOCs referenced throughout multiple steps in the investigation.
• Focused indicator reporting. With the IOC manager's relevance-filtering instruction, the model stopped dumping entire registries into its responses. Indicator lists became proportional to the current step's scope rather than the investigation's total history.

Discussion

Why Extraction Without the IOC Prompt Adjustment Fails

A natural question is why input-side extraction alone does not work. If the prompt already contains a symbolic reference instead of a raw email address, why does the model still generate raw values in its output?

The answer lies in how LLMs process context. The model has access to the full prompt, including sections where the actual IOC value may still appear, task-specific inputs, quoted alert descriptions, or the registry itself. More fundamentally, the model's training distribution contains overwhelmingly more examples of raw IOC values than of symbolic reference systems. Without explicit instruction, the model defaults to the more familiar pattern.

This finding has a broader implication for LLM-based agent design: transforming the input is necessary but not sufficient when you need the model to adopt a non-default output convention. Explicit behavioral instruction, the IOC prompt adjustment, bridges the gap.

Limitations

Our evaluation has several limitations worth noting. All traces were drawn from a single investigation type (phishing via specific security tools). While the IOC types encountered are representative of broader security operations, there are additional evaluations to be done.

The evaluation was conducted with a single model (chatGPT4.1). Different models may exhibit different compliance characteristics, and the prompt mechanism may need tuning for models with different instruction-following tendencies.

Finally, our compliance metric is binary - a response either uses references correctly or it does not. A more granular metric could capture partial compliance and might reveal subtler performance trends across model versions or investigation complexities.

‍

Conclusion

We presented an IOC indexing system for AI-driven security investigations that addresses three interrelated problems: token bloat from repeated raw indicator values, inconsistent IOC tracking across investigation steps, and structural fragility in model-generated structured outputs.

The system combines automated IOC extraction with symbolic reference assignment, explicit behavioral guidance through prompt engineering, and input preprocessing to handle malformed tool output. Across 100 evaluation runs on 10 production investigation traces, the full system achieved 100% JSON validity and 100% IOC reference compliance, up from approximately 80% and 0% respectively at baseline.

The key insight is that managing IOCs in the context of LLM-based agents requires intervention at both the input and output stages. Extraction and indexing normalize the input, but only explicit prompt-level guidance ensures the model adopts the reference convention in its generated output. Neither component alone is sufficient; together, they eliminate the problem entirely.

As SOC automation platforms handle increasingly complex, multi-step investigations, structured approaches to managing the information that flows through the agent's context window become essential. IOC indexing is one instance of a more general pattern: giving the agent a well-organized working memory that scales with investigation complexity rather than against it.

‍

Abstract & Data Summary

We gathered and manually annotated a dataset of 196 hard triage decisions from real-world security investigations, covering a wide range of outcomes, including benign, malicious, and false positives. After cleaning the dataset by removing mock runs and cases with missing information or incorrect workflow execution, the remaining 163 examples were grouped into use case categories to form a high-quality cohort. We then evaluated LLMs on the dataset overall and per use-case category and found that Gemini 3 Pro performs best overall, though the best LLM varies by use case category. 

Model performance by use case category:

If you’d like to understand our full research methodology, read on.

*Note: since this blog was authored, several new model families have been released. While the results have remained broadly stable, particularly among the best and worst performers, updated research may be required for a nuanced understanding of the performance differences amongst the rest.

Data Collection

The dataset was constructed from security investigations from eight US-based customers.The evaluation is conducted in a secure, federated way, without mixing customer data, only reporting summary statistics from each customer tenant.

To create a challenging evaluation, we over-weighted cases in which the analyst dis-agreed with the model - so the error rate is inflated here.

The investigations were conducted automatically according to predefined, customer-specific workflows, each of which contained at least one triage decision node. A triage decision node is a decision point within a workflow, where an LLM chooses a decision from among a list of provided decision options, given the information that was gathered in the workflow up until that point.  

At each decision node, the LLM used in production selected a classification decision from a list of workflow-specific decision options and provided the reasoning for its decision, based on a summary of the steps completed until that point in the investigation.

For each investigation containing at least one decision node, we collected the following information from production session logs:

• A summary of the workflow steps up until the decision node, including tool name, step description, and step outputs
• Organization-specific knowledge, written by the customer and containing a title, description, and data
• The set of available decision options at the decision node
• The model's selected decision in production, as well as the reasoning and detailed reasoning for the decision
• The decision option selected by the customer
• Feedback text written by the customer for the decision

Here is an example workflow diagram:

Quality Control

An expert cybersecurity analyst annotated the 196 decision examples with reasoning tags to explain the production and customer decisions, and label whether disagreements are explained by an analyst-error, mistaken reasoning by the AI or missing data / steps in the workflow. 

Examples tagged with "Workflow ran correctly but missing information" or "Workflow ran incorrectly" were removed from the dataset. Two additional examples with the use case titled "Workshop" were removed, as these were mock runs. For the remaining examples, the workflow ran correctly and was not missing information.

Triage Decision Distribution

By Label

Across the filtered dataset, the workflows contained 27 distinct normalized decision labels, which we grouped into the following buckets: False Positive, True Positive, Requires Review, and Other. The distribution of the labels is shown below: 

The final evaluation dataset contains data from eight customers. The table below shows the number of annotated decision examples per customer and the tools used in each environment.

Use Case Distribution

We consolidated the use cases into 3 categories to consolidate our findings. Below is the map from the consolidated categories to the original use cases, as well as the distribution of the dataset over the consolidated categories. 

‍

Confusion Matrix

Below is a confusion matrix between the expert analyst annotations and the recommendations our system makes. We prompt the models to be careful and escalate when they are not sure. 

Results

Over all use cases (including those without a use case name), Gemini 3 Pro had the highest performance at 74.8%, with GPT-4.1 and Opus 4.5 tied for second.

Phishing Results:

On the phishing use cases, Gemini 3 Pro performed the best, followed by Opus 4.5. 

Account Takeover Results:

Sonnet 4 and GPT-4.1 were tied for best on the account takeover use cases. 

‍

Network Results:

Opus 4.5 and GPT-4.1 were tied for best on the network use cases.

‍

Conclusion & Recommendation

We gathered and annotated 163 triage decisions from real-world security investigations. We characterized the use case distribution, and grouped the use cases according to common categories. We then benchmarked large language models across each use case category and the full dataset. We found that Gemini 3 Pro performs best overall. Per use case category, Gemini 3 Pro gives the best performance on phishing, Sonnet 4 and GPT-4.1 are tied for best on account takeover, and Opus 4.5 and GPT-4.1 are tied for best on network. Based on our results, we recommend that security teams test models for different scenarios to find the solution that works best for their use case, different models are good at different things and the only way to know which model works best for your use-cases it to run formal evaluation - or, you can trust us! Our research team in Legion is constantly evaluating new models and improvements to our triage pipelines.

‍

The security industry spent years debating when attackers would gain capabilities once out of reach — nation-state-level offensive tooling, zero-day discovery at scale, exploits built and iterated in minutes.

That gap was real. And it gave organizations the impression that the decision about which AI to bring into security operations, and how to do it right, could wait until the picture was clearer.

Mythos ended that assumption. 

Not because of the model's size or strength, but because by the time Anthropic announced it, Mythos had already found thousands of high-severity vulnerabilities across every major operating system and browser in use today, without being told where to look. The decision not to release is the signal everyone was looking for.

That changes the implementation question. It was never acceptable to deploy AI badly in the SOC. Now it's not acceptable to deploy it slowly either. The organizations that will come out on top in the next 12 months are the ones that move fast and get it right, and most of the industry is about to discover that those aren't the same thing.

Level set: defenders have always been behind

The average breach lifecycle was already 258 days before AI-assisted attacks became the norm. This has nothing to do with the capabilities of analysts. Human-speed defense against machine-speed offense was always a losing equation.

Mythos-class models will almost certainly expand this breach lifecycle delta. 

Most Implementations Are Getting It Wrong

87% of organizations experienced an AI-driven cyberattack in the past year. Security teams know they need AI. Most are already moving. But most implementations are failing for the same reason, and it is not the technology. It is a missing critical datapoint.

You. The context that shapes your business.

Most AI SOC tools treat every organization as interchangeable. They integrate with your SIEM, your EDR, your threat intel platforms, and assume that is enough. It is not. What determines whether AI actually works in your environment has nothing to do with the list of integrations. It is the organizational context that no integration can capture.

How is your organization structured? Where does data actually live versus where it is supposed to live? Who owns what, and how does that map to investigation and response when something goes wrong? How do escalation paths work in practice, not on paper? And critically, how do you enable the business without interrupting it?

The difference shows up clearly in practice. A heavily regulated enterprise running investigations across proprietary internal platforms looks nothing like a technology company. The organizational context that shapes every investigation, every escalation decision, and every response action is invisible to a system that only sees tool outputs. 

Closing that gap is the foundational requirement that most implementations skip entirely.

Org Context Is Not a One-Time Setup

This is where most implementations fail, even when they start well.

Organizational context is not a configuration you complete on day one. Your organization is a living thing. Teams change. Tools get added. Processes evolve. New subsidiaries appear. Risk posture shifts with every acquisition, every regulatory update, every new product line the business launches. 

An AI system that ingested your context six months ago and stopped learning is already drifting from your reality. It is making decisions based on an organization that no longer exists.

The right model is not a one-off ingestion. It is a continuous learning system that stays embedded in how your organization actually operates, tracks how investigations unfold, incorporates analyst feedback, and updates its understanding as your environment changes. 

Not a snapshot. 

A persistent model of your specific organization that evolves with it.

What Good Implementation Actually Looks Like

First, AI systems needs to understand how your organization actually operates. Not how it is documented, but how investigations really unfold, where data actually lives, and how decisions get made under pressure. The gap between what is written down and what actually happens is where most AI systems fail.

Second, that understanding cannot be static. Organizations change constantly. New teams, new tools, new processes, new risk priorities. Any system that relies on a snapshot of your environment will drift from reality and degrade over time. The AI working in your environment needs to keep learning it, not just learn it once.

Third, it needs to operate within that context, not around it. Producing technically correct outputs is not enough. The system needs to produce outcomes that are actionable within your organization as it exists today. That means working within your existing workflows, tools, and constraints without asking you to change how you operate to accommodate it.

That is the standard. Systems built around this model behave differently from the start. They do not ask organizations to adapt to them. They adapt to the organization. That distinction is where most implementations succeed or fail, and it is where the industry is slowly converging.

The Only Durable Path

The organizations getting AI right in the SOC aren't the ones with the longest integration lists or the biggest models. They're the ones that treated organizational context as the foundation rather than the afterthought, and built systems that keep learning their environment rather than freezing it in place on day one.

That is a harder implementation. It requires more from the vendor and more from the buyer. But Mythos made the timeline for getting there non-negotiable. The organizations that move fast on the wrong implementation will spend the next year rebuilding. The ones that move slowly on the right one will spend it exposed. The only durable path is moving quickly on the version that actually holds up. Systems built on continuous organizational context, deployed now rather than after the next incident, force the question.

The gap that used to buy time for deliberation is gone. What's left is the quality of the decision you make in its absence.

Anthropic was right (and responsible) to release Mythos first to cybersecurity researchers and a select group of organizations through Project Glasswing. It is a genuinely remarkable model. And the security community should take it seriously. What is available to defenders today will be in the hands of attackers in a few months. That window is closing fast.

Mythos raises the ceiling on what AI can do in cybersecurity tasks. It discovers zero-day vulnerabilities in codebases that previous models could not find. It reverse-engineers complex systems. It constructs sophisticated, multi-path exploits at scale. The capabilities that were previously accessible only to well-funded nation-state actors can now be replicated by a far broader set of threat actors. No longer do you need teams of expert reverse engineers and months of reconnaissance.

The threat landscape is structurally shifting. We will be determined by our ability to shift our defense in kind. Quickly.

Where AI in defense needs to go first

The industry is converging, rightly, on vulnerability research and remediation as the priority. Scanning your own codebase with the same class of models that attackers are using is a clear first step. In many cases, defenders actually have an asymmetric advantage here, as we have better access to our own code than attackers do.

The harder problem is remediation. We already carry significant backlogs of unresolved, sometimes exploitable, vulnerabilities. Unlike an attacker who has nothing to lose, defenders cannot afford mistakes. Our systems are in production. Downtime has real costs. The asymmetry of attacker agility versus defender accountability is where the gap widens.

AI-assisted vulnerability remediation at scale is necessary. But it is not a solved problem, and any honest assessment of the landscape has to acknowledge that.

What this means for security operations

The idea of static detections designed to discover dynamic adversaries is fundamentally misguided. The future is better trip wires and an assume-breach mentality.

For SOC teams, the implications are direct. The scale and complexity of attacks is accelerating. We should expect a higher volume of sophisticated attacks that actively evade detection, that do not conform to known signatures or behavioral patterns, and that are designed from the ground up to stay invisible.

This breaks the model that most SOC programs are built on. The idea of maintaining a library of static detections to catch dynamic adversaries has always had limits. Those limits are now being exposed in real time.

What we need instead is the ability to detect a high volume of low-fidelity signals, such as anomalies in endpoint behavior, data access patterns, email activity, network flows, and identity. This requires teams to investigate each one as if it were the leading edge of a sophisticated breach. Not because every alert is a nation-state intrusion, but rather, we should expect that a higher percentage now may be.

The question is no longer whether to adopt AI in security operations. This is clearly needed. We cannot scale defenses solely on human labor. 

The question is how to do it in a way that actually works inside the operational reality that security teams live in.

The real challenge is operational reality

Enterprises have legacy and custom tools, established processes, compliance and audit requirements, escalation paths, and oversight obligations that are not optional. AI cannot simply replace this infrastructure. It has to work within it.

You cannot properly scale your defenses without giving AI access to your organizational context, including your tools, your processes, your detection logic, and your escalation criteria. AI agents need to be able to investigate with the consistency and rigor of an incredible IR analyst, operate transparently, and support human oversight at the points where it matters.

This is precisely what we built Legion to do: meet organizations where they are. Our platform learns your existing tools, processes and context and makes them accessible to the latest frontier models (now Mythos, and every model in the future). From that we create structured, repeatable workflows where consistency is required or fully agentic investigations that require depth and judgment. Every action is auditable. Human-in-the-loop controls are configurable. And the system integrates across your entire stack.

My conclusion - Assume breach, investigate everything, build for the attacker that has already found the vulnerability you have not patched yet, and is using Mythos-level models to stay ahead of your detections.

Picture a senior analyst mid-investigation. Eight browser tabs open across CrowdStrike, VirusTotal, Defender, and Microsoft Entra. She's running a hunting query in one window, checking an IP reputation score in another. And somewhere in between, she's documenting. Taking screenshots, copying log entries into a case note, capturing context before it slips away.

This is the job. Investigations today aren't just about finding the threat. They're about moving across tools, pulling together evidence from a dozen different sources, and building a record that another analyst, or an auditor, or a manager, can actually follow. The documentation isn't a distraction from the work. It is part of the work.

Everyone in security has lived that. 

Which raises a question that's been easy to ignore until now: if we wouldn't accept an analyst who said "trust me, I looked at it"- why are we accepting that from AI agents?

Evidence Has Always Been the Standard

The reason SOC analysts document isn't distrust. It's precision. A good investigation has always meant showing your work. The summary an analyst writes is their claim, the insight they've drawn from what they saw. The screenshot is the fact. Undisputable evidence, captured at the moment of discovery. Together they tell the full story: here is what I found, and here is the proof.

Evidence gathering has always been a core part of the job. Screenshots and logs aren't bureaucratic overhead. They're how you distinguish signal from noise, how you close out audit findings, how you hand off a case without losing context.

You Wouldn't Accept "Trust Me" From an Analyst. Stop Accepting It From AI

We hold human analysts to a clear standard. When an analyst closes a case, we expect to see their work. The exact screen they reviewed, the exact query they ran, the exact result that informed their decision. A summary of what they found is a claim. The screenshot is the proof.

We should hold AI agents to the same standard.

Today, most AI SOC give you a verdict and a reason. The agent processed the alert, evaluated the indicators, and concluded it was malicious. But if you ask what it actually saw, you're directed to API logs and structured JSON responses. That's not evidence. That's a reconstruction built after the fact, from data that was never meant to be read by a human auditor in the first place.

The gap between what an AI agent did and what you can actually verify is where hallucination risk lives. A summary can sound confident and still be wrong. Without visual evidence captured at the moment of the decision, you have no way to know what the system actually encountered.

Legion operates differently. Instead of calling APIs, Legion navigates your source systems directly through the browser, the same way a human analyst would. It opens the actual system, reads the actual screen, and captures a screenshot of exactly what it sees at every step. The summary is the claim. The screenshot is the fact.

That's the standard we believe AI investigations should meet. And it's the only architecture that meets it. 

How Legion Automates Evidence Gathering

Legion Evidence Gathering captures visual proof of every action Legion takes as it navigates your source systems, automatically, in real time.

Take a malware investigation spanning CrowdStrike, VirusTotal, and Defender. Legion opens the originating ticket, reads the case, and begins investigating. As it moves through each tool, it takes a screenshot at every step. The CrowdStrike detection page as it appeared. The VirusTotal result in context. The Defender hunting query and its output. Every interface, exactly as Legion saw it.

By the time an analyst opens the case, the full evidence gallery is already there. Screenshots organized sequentially, labeled by tool, timestamped, and ready to review. Not just a summary. Not just a log. The complete picture: the analysis and the visual evidence behind every conclusion.

And it stays there. Every investigation Legion runs is stored and searchable. When an auditor asks a question, when a peer analyst picks up a handoff, when someone needs to understand why a decision was made, you go back to the session and everything is right there. Every step. Every screen. Nothing reconstructed. Nothing missing.

Different alert types. Different toolchains. The same complete evidence gallery, every time.

This Is What Accountable AI Looks Like

We've always known what a good investigation looks like. You show your work. You back your conclusions with evidence. You leave a record that someone else can follow. Legion applies that same standard to every automated investigation it runs, without exception and without manual effort. The bar doesn't move because the analyst is an AI. It stays exactly where it's always been.

See Legion Evidence Gathering in action. Request a Demo

SOC investigations range widely. Some are highly repeatable: every step defined, every decision documented. These work well and can be fully automated. But some investigations eventually reach a point where that breaks down: where the next step depends on what you just found, and the judgment and intuition to know what it means.

You can see it clearly the moment you try to write it down. Some processes flow neatly from start to finish. But as soon as you move into more complex investigations, the cracks appear. You find yourself pulled into a spiral of edge cases, tool variations, and fallback paths. You add branches. Then branches on branches. And after all that effort, you almost always end up in the same place: where no rule applies, and only judgment, reasoning and intuition can take you further.

The Part You Can Never Quite Capture

SOC investigations don't all look the same. Some are fully deterministic: a user notification when an outgoing email gets blocked, no reasoning required. For these, consistency matters. The same steps, the same outcome, every time. Others are the opposite: novel threats with no fixed path, no known pattern, where only experience, intuition, and judgment can tell you what to do next. And many fall somewhere in between, where you start with structure and hit a point where judgment has to take over.

But even those flows have a ceiling. Take a phishing investigation. You can document the triage steps pretty cleanly: check the sender, analyze the headers, detonate the attachment, check the URLs. That part is routine and capturable. But the moment you find something suspicious, the investigation shifts. Now you need to reason about scope: is this part of a campaign, and who else was hit? That question has no fixed answer. You might search for other emails with the same subject, but any decent campaign will vary the lures across targets, changing subjects, sender names, and payload links to evade detection. You cannot match on a single field and call it done. You need to iterate: follow one thread, see what it reveals, adjust your search, go again. You are reading the environment in real time, making judgment calls at every step based on what the last one uncovered.

Those judgment points show up on every shift, on every alert that goes beyond the routine. Someone has to reason through them in the moment, with whatever context they have, under whatever pressure exists right now. Until 3am. Until a less experienced analyst picks it up. Until alert volume means there simply isn't time to think it through properly.

That reasoning is not pre-programmed. It emerges from the finding itself. It is what a senior analyst does instinctively, and until now there has been no way to replicate it at scale. Legion Investigator is built for that moment.

Your Environment. Your Logic. Your Investigator.

Legion Investigator is a goal-oriented AI agent that sits inside your investigation workflow at exactly the moments where reasoning takes over from execution, extending Legion's coverage across the full spectrum of SOC investigations, from fully deterministic workflows to complex open-ended investigations. You define its goal, you choose which tools and actions it is permitted to use, and you decide where it acts autonomously and where it checks in first. 

Which category a given investigation falls into is sometimes obvious. But often it is a deliberate choice, one that should be yours to make based on your team's needs, your risk tolerance, and how much consistency versus flexibility the situation calls for. Where on that spectrum each investigation runs is yours to decide. Every boundary is one you set in advance and can trust will be respected. This is what makes Investigator the kind of AI enterprises can actually adopt: not just powerful, but designed from the ground up to operate within your constraints, your processes, and your level of trust.

Most AI SOC tools bring their own model of how investigations should work. Legion Investigator learns from how yours actually do. It builds its understanding from your team's recorded investigation sessions, the decisions they make, the paths they take, and the patterns that emerge across real cases in your environment. Over time, Legion builds a structured knowledge base specific to your organization, capturing your processes, your tooling, and your team's accumulated expertise. That knowledge is not just stored. It is actively used to improve your captured workflows and feeds directly into how Investigator reasons, prioritizes, and investigates.

And when we say your tools, we mean all of them. Legion Investigator works the way your analysts work, through the browser, with no integrations and no APIs required. Your SIEM, your EDR, your threat intelligence platforms, your homegrown applications, your legacy dashboards, your on-prem and cloud environments. You don’t rebuild your stack to fit the tool. The tool fits your stack.

The way it works reflects how investigations actually flow. An investigation might start in your SIEM with a set of routine queries, structured, reliable, repeatable. But when it reaches one of those decision points, you hand off to an Investigator with a goal: find the scope of breach, enrich the full context of what we have so far, identify what else was impacted across endpoints and cloud assets.

The Investigator takes that goal and works toward achieving it. It invokes the relevant tools, interprets what comes back, recalculates what to do next, and invokes again. It keeps going, step by step, until the goal is met. Not a single tool call with a result handed back to you. A full reasoning loop that runs until the work is done, across your security tools, your homegrown applications, and any AI agents already running in your environment. Investigator acts as the orchestrator, pulling in whatever is needed to get there.

Multiple Investigators can work together across a single investigation. One handles enrichment. Another determines scope of breach. A third drives containment based on what was actually found, not what was anticipated when the playbook was written.

And because trust matters, Investigator operates within guardrails. It works only with the tools and actions it’s been given permission to use. For anything higher risk, it asks before acting. You stay in control by setting the boundaries in advance and knowing they’ll be respected.

What This Changes

Legion Investigator opens up three things that weren't possible before.

Pick up where deterministic processes end 

For investigations where you have structured steps, you can now embed an Investigator at exactly the points where structure runs out. The routine parts stay routine.The investigator reasons further, and by the time you step in, the groundwork is already done.

Handle your long tail of alerts

For the long tail of investigations where you never had a well-defined flow to begin with, you can now hand them off end to end. The Investigator handles enrichment before you even open the case, drives containment the moment scope is confirmed, and picks up every judgment point in between.  Give the Investigator the goal, set the guardrails, and let it run. No playbook required.

Every investigation, regardless of how well-defined it is, can now be handled with the depth of your best analyst, on every alert, on every shift. And for the first time, you control where on that spectrum each investigation runs. More structure where consistency matters. More autonomy where judgment, experience, and intuition are required. The balance is yours to set, and yours to change.

This is not about replacing analysts. It never was. There will always be moments that require human judgment, experience, and instinct, and no AI should pretend otherwise. What changes is everything around those moments. The analyst becomes the commander: setting goals, defining boundaries, sending investigators out into the environment to gather, reason, and report back. The calls that matter stay with you. The work that surrounds them no longer has to. Not because we built a smarter AI. Because we built one that learned from you.

‍