DeepMind co-founder warns AI hacking incidents are ‘early warning shots’

Google DeepMind co-founder Demis Hassabis on Wednesday warned that hacking incidents involving artificial intelligence are “early warning shots,” urging companies and governments to strengthen security as autonomous systems become more powerful.

Speaking at the International AI Summit 2026, Hassabis said AI agents capable of acting on users’ behalf increase both the technology’s usefulness and its risks.

“But also that means that increases also the risks involved with what those systems might do, and we've seen that recently earlier this year with some of these hacking incidents. I think these are early warning shots for us,” he said.

“The systems aren't that powerful yet, so the incidents aren't that major,” he added, warning that incidents could become more serious over the next few years without additional security efforts.

Hassabis said ensuring that systems follow their assigned goals and remain within established safeguards becomes harder as they gain autonomy.

“And that gets increasingly difficult the more automation there is in the systems,” he said.

He also warned that AI “can also be repurposed by bad actors for harmful ends.”

Hassabis called for testing systems in secure environments, initially disconnected from the internet, and drawing on containment practices from biology and nuclear technology.

“I think there's a lot of lessons that can be learned about containment systems, and I think we need to start taking that seriously and bring experts in from those areas,” he said.

He advocated giving trusted defenders, including banks and governments, access to advanced AI tools before wider release to identify and repair software vulnerabilities before attackers could exploit them.

“So we've got to kind of asymmetrically power up our defensive side to keep ahead of the attackers,” he said.

Hassabis also called for an international standards body capable of adapting to technological developments.

“It's going to affect everyone in the world,” he said, urging safeguards that reflect emerging risks while preserving innovation.

“If you legislate for something three years ago, it will be out of date today, and actually probably won't even address what the real problems are,” he said.

He also called for “10 times or more resources” for AI safety and security, including research and computing capacity.