Google DeepMind chairman Demis Hassabis has let out a warning at an AI summit in Dublin that recent hacks involving autonomous AI agents could be early signals of bigger security threats. He urged companies and governments to strengthen their defenses before AI technology becomes more capable.
DeepMind co-founder and Alphabet chief scientist Hassabis struck a balance between optimism about AI’s scientific potential and warnings about its security risks at the International AI Summit.
He cited hacks by autonomous AI agents that caused no serious damage but exposed underlying security risks, stressing that these issues should not be ignored, as the systems were “not healthy yet.”
Hassabis said the technology was just not yet powerful enough for those incidents to have inflicted major harm.
Deepmind chair wants defenders armed first
Hassabis said security teams need an advantage before AI tools reach the public. “I think one of the big things we need to do is asymmetrically power the defenders, trust the defenders,” he stated. He added that the same systems should help find and patch software vulnerabilities before release.
Hassabis urged companies and governments to create secure sandboxes and test autonomous agents thoroughly before release. He also suggested drawing on experts in nuclear technology and biology, fields with experience managing dangerous systems, to develop ways to contain AI risks.
The DeepMind co-founder also warned about misalignment, the risk that AI agents may stray from their operators’ goals. “It increases the risks involved with what those systems might do, we’ve seen that recently,” he said at the summit.
Hassabis proposed an international standards body to establish safeguards for advanced AI, while stressing that any framework must keep pace with the technology without stifling research. He also proposed a certification process based off potential risks that evolves as new capabilities and vulnerabilities emerge.
Microsoft’s President Brad Smith claimed 2026 was a “sobering year” for AI safety and urged global regulators to stay abreast of AI’s development. He added that no company or technology should be above the law, as previously reported by Cryptopolitan.
A Nobel laureate’s optimism
Hassabis won the 2024 Nobel Prize in Chemistry alongside John Jumper for developing an AI model that predicts the structures of nearly all 200 million known proteins. The technology is already helping researchers study antibiotic resistance and design plastic-degrading enzymes, according to the Royal Swedish Academy of Sciences.
His track record regarding his use of AI guided his statements in Dublin, where Hassabis described AI as a dual-use technology that can help treat disease and perform a lot of exploits, but which can also be exploited by bad actors.
He argued that security measures should protect AI’s potential benefits in the long-term, and not completely hinder them.
The smartest crypto minds already read our newsletter. Want in? Join them.