"DeepMind": Current AI incidents are early warnings of what may happen soon

Demis Hassabis, chairman of Google DeepMind, warned that AI systems could cause major security incidents within the next two years unless companies and governments take additional safety measures. He described recent incidents involving AI agents as "early warnings," noting that the limited capabilities of current systems have prevented more serious accidents.

Demis Hassabis, chairman of Google DeepMind, warned that AI systems could cause major security incidents within the next two years unless companies and governments take additional safety measures. Hassabis made the remarks at an AI conference in Dublin, Ireland, where he described recent security incidents involving AI agents as "early warnings." He noted that the limited capabilities of current systems have so far prevented more serious accidents, but cautioned that these capabilities may increase over the next year or two. According to Reuters, Hassabis's comments come amid growing concerns about AI safety, following incidents including breaches of institutional networks by AI agents, as well as Anthropic revealing that one of its models filed a false police report about a murder. Hassabis called for adopting safety standards from sensitive sectors, such as biological research and nuclear energy, and applying similar practices to the development and testing of AI systems. He also suggested that establishing an international body to set safety standards for this technology may become necessary.