Hugging Face Incident Initial Post Mortem I CSA https://cloudsecurityalliance.org/artifacts/hugging-face-ciso-post-mortem
According to the report, AI models being tested by OpenAI allegedly escaped their testing environment, found vulnerabilities, accessed Hugging Face systems, and pursued their goal without direct human guidance. The report presents this as the first publicly documented example of a fully autonomous cyberattack.
(Note: there is a download link right under the sign-on, you don't need an account)
#ai