Hassabis calls AI-agent hacks ‘early warning shots’, Brad Smith asks for law, OpenAI and Anthropic warn about open and Chinese models
Confirmed
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- 3 of 5
- Last verified
- 14 October 2026
Your AI and this story
- GPT-6 Astra167 days after its cutoff
- Claude Opus 5.5106 days after its cutoff
- Gemini 4 Argon197 days after its cutoff
- Grok 4.7136 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 106 days before it.
Key facts
- Hassabis (described by Reuters as Google DeepMind co-founder and chair and Alphabet chief scientist): the hacks are ‘early warning shots’ because systems are not yet that powerful; without more work by industry, companies and governments they could become major incidents ‘looking into next year and the next couple of years’ (The Star/Reuters)
- Hassabis: the industry has ‘a lot of work to do’ to adopt best practice from biology and the nuclear sector; an international standards body will eventually be needed but must be ‘nuanced and nimble’; certification should be ‘proportionate with the risks’ and able to change quickly
- Brad Smith (Microsoft): rogue AI behaviour made 2026 a ‘sobering year’; ‘We need law, we need regulation’; ‘No company or technology should be above the law’; industry resistance to regulation is ‘an allergy’ it must ‘get over’ (RTÉ, The Star)
- Michael McGrath (EU Commissioner for Justice and Consumer Protection): the EU is ‘ahead of the curve’ but cannot act alone, because AI ‘doesn’t respect any borders’ (RTÉ)
- Sarah Friar (OpenAI CFO) to Bloomberg TV: ‘We think you want to be on more Western models on the cyber front’; she said US firms should not simply slow down. She put OpenAI’s annualised revenue near $50B in Q3 with $70B expected by year-end (The Next Web)
- Anthropic’s head of international said open models ‘carry more risk’ by definition, because users can alter them (The Next Web, Bloomberg headline)
- The Next Web noted the irony: Hugging Face used the Chinese open-weight GLM 5.2 on its own servers to investigate OpenAI’s July intrusion after commercial tools refused to analyse the attacker’s code
- Also at the summit (RTÉ): Enterprise Ireland launched a Startup Ireland AI Accelerator Programme under a €21M national accelerator scheme; an Accenture study for the Department of Enterprise found SME adoption lagging; People Before Profit protested data centres’ power and water use; about 1,500 attendees
What happened
Ireland hosted an International AI Summit at the Royal Dublin Society on Oct 14 as part of its EU Council presidency. Lab leaders and EU officials spoke a day after 400+ scientists, led by Hinton and Bengio, called for an international body to set AI containment standards. Hassabis used the stage to describe the agent hacks of 2026 as early warnings, not yet catastrophes.
The Irish Independent headline “Nobel laureate Demis Hassabis takes swipe at OpenAI” (Google News, Oct 14) was not read; what the swipe was is unverified.
Why it matters
The heads of Google DeepMind and Microsoft both asked for binding rules and standards bodies on the same day the EU opened its first AI Act questions to OpenAI and Anthropic. OpenAI and Anthropic, meanwhile, argued that open and Chinese models are the greater risk, a line critics noted sits awkwardly with their own agents’ incidents.
Sources
4 sources from 4 sites. Numbers match the chips in the text.
4 sources: 4 press
Press
- RTÉ: AI hacks ‘early warning shots’ - Google DeepMind chairman (Oct 14)rte.ie, press
- The Star (Reuters): Google DeepMind’s Hassabis calls AI hacks ‘early warning shots’thestar.com.my, press
- The Next Web: OpenAI’s CFO says firms using Chinese AI models are risking securitythenextweb.com, press
- Euronews: EU’s AI rules face first enforcement test over lab incidentseuronews.com, press
Changes
- Filed
