Trail of Bits AI/ML Security
Security assessments of AI systems end to end, including model capability evaluations.
Offered by Trail of BitsRed teaming and evaluation
About Trail of Bits AI/ML Security
Trail of Bits offers an AI/ML Security service that assesses AI/ML systems end to end. The firm states that AI failure modes live in the seams between disciplines, so it staffs every engagement across ML, application security, systems, and cryptography.
Trail of Bits was founded in 2012 and states that it publishes its work publicly. The service's Model Capabilities Evaluation offering helps organizations measure and validate the capabilities of the AI models their systems employ, both first- and third-party. The firm also offers custom training courses on AI failure modes, adversarial attacks, and AI safety.
Drafted from trailofbits.com · not yet confirmed by Trail of Bits
What they do
Trail of Bits reviews AI/ML systems from training data and MLOps pipelines to model artifacts, inference hardware and deployed agents. Engagements include pipeline assessments, AI risk assessments with threat modeling, adversarial probing of deployed models and agents (prompt injection, data exfiltration, inference manipulation, capability misuse), and evaluations of models' offensive and defensive cyber capabilities benchmarked against experts, tools and novices.
Each engagement is staffed with cryptographers, application security engineers and systems specialists alongside its ML team, and ships fixes such as Semgrep and CodeQL rules and fuzzing and evaluation harnesses. It also runs custom training courses on AI failure modes, adversarial attacks and AI safety.
How they work
- How you engage them
- Project
Reviews
Be the first to review it
AnonymousKnow a team weighing Trail of Bits AI/ML Security? Send them this page.