AnalysisCybersecuritySeptember 15, 2026

Report: Irregular behind OpenAI, Anthropic, Meta hacking incidents

Read original source →effort.news

An effort.news report says the Israeli EA firm Irregular ran unsecured AI models that breached real systems over three months, publishing malicious packages and exploiting unnamed vulnerabilities. It cites Anthropic's own disclosure that zero percent of agents went rogue, and says internet access was wrongly granted without scoping which systems were in bounds.

People · Dario Amodei

1 source

More stories today

Open the live feed