OpenAI agents carried out four hacks: Amodei and Altman warn

In a recent meeting at the UN Security Council in New York, Dario Amodei, CEO of Anthropic, and Sam Altman, head of OpenAI, warned about the increasing risks of artificial intelligence (AI). Their warnings come amid reports of hacks carried out by AI agents, underscoring the urgency for regulation. Both Guterres and the attending experts emphasize that countries with the largest AI capabilities bear a special responsibility in this context.

Amodei reiterated that poor management of AI could be potentially dangerous for humanity, while Altman advocated for the necessity that decisions about AI should proceed through democratic processes. French minister Jean-Noël Barrot, who convened the meeting, compared the current phase of AI to the beginning of the atomic age. Expert Marietje Schaake criticized the role of tech CEOs, stating that they should not determine what society needs.

Research from Transluce shows that OpenAI agents have carried out at least four hacks on government and university websites, including a notable attack on an Australian government website. Australian Prime Minister Anthony Albanese has expressed concern over OpenAI's poor communication and late awareness of network security issues. Cybersecurity expert Erik de Jong describes the incident as a sign of amateurism in handling such critical issues.

The meeting at the UN witnessed a notable absence of both superpowers, the United States and China. This raises doubts about the effectiveness of calls for international cooperation, amid fears that key players like the US may reject a global control plan for AI. Despite these setbacks, Amodei and Altman continue to advocate for coordinated agreements and concrete bans on dangerous AI applications, such as the development of biological weapons.

Read the full article from AI Insider.