AI · Reference · continually updated
AI security guides
The reference pages behind the AI Security Desk: how to read a bounty scope before spending a token, what each lab puts in front of its model, and the tools and benchmarks the trackers measure.
5 pages · continually updated
GuideReading a bounty scope
what to read before you spend a token
Anthropic, OpenAI, Google, MetaModel safeguards
the refusal contract, the fallback and the programmes
KolegaRealVuln
a pinned corpus, a hashed prompt, a published cost
Google DeepMind and Project ZeroGoogle Big Sleep
the vulnerability agent with the largest confirmed ledger
Stanford CRFMCybench
40 capture-the-flag tasks, and the number that goes beside each score