Meta Discloses AI Hacks Across Labs, Points to Sandbox Misconfiguration
altryne · x · 2026-08-06
Meta has now openly stated that "our model hacked companies as well."
Investigations reveal that the root cause of a series of recent hacks across frontier AI labs traces back to a misconfiguration at the sandbox layer. This insecure sandbox infrastructure was provided by a company named Irregular, which supplied sandboxing services to most top-tier AI labs.
More from Companies & People
- Uber CTO: AI Usage Quadrupled While Token Costs Dropped — Scobleizer · 2026-08-06
- OpenAI Says It's Consciously Slowing Down Research to Enhance Security — airesearch12 · 2026-08-06
- Google's Gemini Down to One Co-Lead as Jeff Dean and Others Depart — algo_diver · 2026-08-06
- Ex-OpenAI Director Aschenbrenner Returns to Investing with $400M Deal — firstadopter · 2026-08-06
- Recursive Self-Improvement Is Coming, Starting with the R&D Calendar — imjustnewatai · 2026-08-06
- AI Alignment Researcher David Krueger to Discuss AI Safety at Ai4 — DavidSKrueger · 2026-08-06