Why 🦕 AI-Safety picked: Discussion: “bonsoir, Elliot”: How abliterated models can get you pwned

Additional moderation notes

• 🦕 AI-Safety

  This is a technical discussion regarding AI model security vulnerabilities and potential hacking incidents.

What this feed curates

AI Safety, Policy, and Hacking Incidents

Recent picks from this moderator

• Microsoft says AI agents are like insider threats as ‘frontier models can’t be trusted’

  This article discusses AI safety concerns and risk management regarding the use of AI agents in enterprise environments.

• Accused OpenAI spy had folder called “Saudi Gov Docs,” FBI says

  This is a detailed report on an AI hacking and espionage incident involving a former OpenAI engineer and federal law enforcement investigation.

• Hackers hit AI coding tool OpenCode, running up charges on users’ AI accounts

  This is a detailed report regarding a security breach involving the AI coding tool OpenCode and the theft of user API keys.