Google's NotebookLM is now the foundation for a new open-source security tool that helps organizations identify AI-specific threats before they become exploitable vulnerabilities. The OWASP Foundation's Threat Advisor is an AI-powered assistant that guides users through threat modeling by asking them to describe their AI system and then conducting an interactive interview to identify relevant risks, assess severity, and recommend mitigations.

The tool represents a significant shift in how organizations approach AI security. Rather than manually navigating hundreds of pages of technical documentation, teams can now use a conversational interface to surface threats relevant to their specific deployment. The recommendations draw on more than 300 pages of OWASP AI security guidance based on frameworks and standards such as MOSAIC.

What AI-Specific Security Risks Is Threat Advisor Designed to Catch?

Organizations deploying AI systems face a growing list of threats that traditional security practices don't address. Threat Advisor helps teams identify and prioritize these emerging risks within their own environment.

• Prompt Injection: Attackers manipulating AI model inputs to bypass safety guidelines or extract sensitive information
• Supply Chain Compromise: Vulnerabilities introduced through third-party models, training data, or dependencies used in AI systems
• Excessive Model Privileges: AI agents granted too much access to systems, data, or external tools without proper controls
• Insecure Agent Deployments: AI systems operating autonomously without adequate monitoring, logging, or human oversight mechanisms

Rob van der Veer, founder of the OWASP AI Exchange, emphasized the tool's practical value in a recent LinkedIn post, stating that Threat Advisor is "the fastest way to understand your exposure and where to focus next".

How Does Threat Advisor Actually Work in Practice?

The tool operates as a guided conversation within NotebookLM's chat interface. Users describe their AI system, and the assistant asks targeted questions to understand the system's architecture, data flows, and operational context. Based on these responses, Threat Advisor maps the organization's deployment against known threat patterns and control frameworks.

Security experts view the tool as particularly valuable for organizations lacking dedicated AI security teams. Seemant Sehgal, founder and CEO of BreachLock, explained that the automated interview can surface relevant controls and map to known threat patterns, though the quality of output depends heavily on input accuracy. "The value of a tool like this depends entirely on what the organization brings to it," Sehgal noted. "But it can only reason about what it's told, and most organizations struggle to articulate their own system boundaries, data flows, and trust assumptions clearly enough to get precise output".

"If you are building or deploying AI, use this to get a high level threat model. It is the fastest way to understand your exposure and where to focus next," said Rob van der Veer.

Rob van der Veer, Founder of the OWASP AI Exchange

Jacob Krell, senior director of secure AI solutions and cybersecurity at Suzu Labs, noted that the tool fills a critical gap for startups and mid-market companies. "Threat Advisor turns that into a guided conversation you can run in a single sitting. For startups and mid-market companies shipping models without a security function, that's a real upgrade from 'we'll deal with it later'".

What Are the Limitations Organizations Should Know About?

While Threat Advisor offers significant value, security experts caution against treating its recommendations as a complete threat model or final decision-making authority. Jeff Williams, founder of OWASP and founder and CTO of Contrast Security, emphasized that the tool is an early prototype and that effective threat modeling requires looking beyond AI-specific risks. "The most serious risks aren't AI-related," Williams explained, noting that longstanding application security issues such as authentication weaknesses, access control failures, and injection flaws remain equally relevant.

Williams also questioned whether an AI assistant can fully capture the architectural and operational context needed for a comprehensive threat model. "It's really all about context," he said, adding that organizations often lack a complete understanding of their own technology stack, data flows, trust boundaries, and threat environment.

"It's a very interesting effort from the volunteers at OWASP, who have been doing a fantastic job at driving AI security forward," said Jeff Williams.

Jeff Williams, Founder of OWASP and Founder and CTO of Contrast Security

Joshua Marpet, senior product security consultant at Finite State, recommended running analyses multiple times with different prompts to help validate and refine results. "Be careful believing it 100%. Not that it's wrong. Just ask the question several ways," Marpet advised.

Another practical limitation: Threat Advisor produces its results as a NotebookLM chat conversation rather than structured data, meaning recommendations cannot be easily imported into enterprise governance, risk, and compliance tools. Someone must manually read the output, extract the findings, and enter them into the organization's risk management system.

Steps to Get the Most Value From Threat Advisor

• Run It Early: Use Threat Advisor before your architecture decisions harden, treating the output as a starting point rather than a final compliance artifact
• Pressure-Test Results: Bring the threat list to your engineering team and validate it against how your system actually runs in production
• Provide Direct Access: Give the AI assistant direct access to your code repository and technical documentation to help answer the tool's questions more accurately
• Validate Output Quality: Organizations that benefit most are those with practitioners who can validate the output, not those hoping the tool will do the thinking for them
• Seek Feedback Loops: Help make the tool better by providing detailed feedback to the Threat Advisor team about your experience and results

Krell emphasized the importance of treating Threat Advisor appropriately within the broader security workflow. "Do not let a NotebookLM conversation become your compliance artifact," he cautioned.

Krell

The emergence of Threat Advisor signals a broader shift in how organizations approach AI security. Rather than treating AI threat modeling as a specialized, resource-intensive process, tools like this are making it more accessible to teams of all sizes. However, experts agree that automation is a starting point, not a replacement for human judgment, architectural understanding, and ongoing security vigilance.