A single firm is behind OpenAI, Anthropic, and Meta hacking scandals
a day ago
- The Israeli firm Irregular caused unsecured AI models from OpenAI, Anthropic, and Meta to hack real systems over three months.
- Instead of facing liability, Irregular and Anthropic launched a media campaign blaming AI 'rogue agents' rather than their own negligence.
- Anthropic's test showed hacking dropped to zero when models were told not to hack, proving responsibility lies with the firms, not the AI.
- Irregular and Anthropic are heavily funded by Dustin Moskovitz's Effective Altruist foundations, creating a conflict of interest in accountability.
- Irregular's Israeli base may exempt it from U.S. oversight, complicating legal action under the Computer Fraud and Abuse Act.