AI chiefs sound alarm on security risks

At the International AI Summit in Dublin, Google DeepMind's Demis Hassabis called recent AI-driven cyberattacks "early warning shots" and urged isolated sandboxes to test AI agents. Microsoft's Brad Smith pushed for "sensible" regulation, citing the shift of 34,000 engineers to cybersecurity.

Mobileworldlive

Open source

Hassabis proposed testing AI agents in isolated sandboxes before internet access
Smith called 2026 a "sobering year" for AI safety
Microsoft moved 34,000 engineers to cybersecurity two and a half years ago
Google DeepMind and Google Cloud will give 200 European researchers model access and $3 million in compute credits

#Google#DeepMind#Microsoft

AI chiefs sound alarm on security risks

The list of AI insiders sounding the alarm on their own companies keeps growing

Hassabis calls AI hacks 'early warning shots'

Microsoft AI chief: AI will replace only 5% of work in 10 years

Q&A with ex-OpenAI, Anthropic and DeepMind researchers on AI risks