Anthropic on Open AI: Amodei Fears Attacker Advantage

Anthropic did not sign an open letter supporting open-weight AI models, despite backing from 77 other companies, including OpenAI and Google. Anthropic CEO Dario Amodei explains his company's position, stating he agrees with much of the letter, but not its claim that open models necessarily help defenders more than attackers. Amodei believes open AI models could empower attackers more. He cites the risk of malicious groups using AI to create bioweapons. While Anthropic has never advocated for banning open-weight models, Amodei says they present a higher risk than closed models because guardrails are difficult to apply and monitor once released. He argues that banning their use by US businesses wouldn't stop bad actors. This highlights ongoing debate about the security implications of open-source AI.