'Early warning shots': Google DeepMind's Demis Hassabis warns AI could trigger major incidents within 2 years

Recent hacking incidents involving artificial intelligence (AI) agents are "early warning shots", and more serious incidents could occur in the coming years if adequate security measures are not implemented, Google DeepMind co-founder and Chair Demis Hassabis warned on Wednesday.

Concerns over the rapidly advancing technology have grown following reports of rogue AI behaviour, including attacks on corporate networks. More recently, Anthropic disclosed that an AI model had submitted a false homicide tip to the police, raising further questions about the potential risks posed by AI systems.

Speaking at an AI summit in Dublin, Hassabis stressed the need for stronger safeguards and security practices to prevent AI-related incidents from escalating. He said there was "a lot of work to do" to introduce the kind of best practices followed in sectors such as biology and nuclear technology to protect critical systems.

Hassabis also called for international cooperation, suggesting that an international standards body would eventually be necessary to establish safeguards for AI systems.

"I think these (hacking incidents) are early warning shots for us that systems aren't that powerful yet, so the incidents aren't that major," Hassabis, who is also chief scientist at Google parent Alphabet, told the summit.

"But they could be (major) looking into next year and the next couple years if we don't do more work as an industry, as individual companies and governments on securing these systems."

AI security needs stronger safeguards

Hassabis's remarks highlight growing concerns about the security risks associated with increasingly capable AI agents. Their ability to perform tasks autonomously has raised questions about how such systems could be misused or behave unexpectedly when operating in digital environments.

He emphasised the need for technology companies, governments and the wider industry to work together to strengthen security measures before AI-related incidents become more serious.

Livemint