Key Notes
- Geoffrey Hinton warns that a major AI-related crisis could affect electricity supplies or financial systems by 2027.
- Controlled industrial security tests have found software vulnerabilities while also revealing model errors that require expert review.
- Governments are calling for independent evaluations and stronger oversight before frontier AI systems are deployed.
Geoffrey Hinton has warned that an AI-related disaster could arrive in late 2026 or 2027, pointing to electricity supplies and financial systems as potential targets. His comparison with Chernobyl raises a difficult question for the AI industry: how much damage could increasingly capable systems cause before their developers or operators can intervene?
The Nobel Prize-winning computer scientist’s warning was reported by The Guardian on November 3. Hinton also suggested that humanity might have been better off without AI. These are his judgments about the technology’s risks, rather than evidence that a particular attack is planned or that a disaster has a scientifically established deadline.
What an AI “Chernobyl” Would Mean
Hinton’s analogy concerns a disruptive technological crisis, with AI potentially helping attackers undermine essential infrastructure. It should not be read as a prediction of a nuclear accident. The central concern is that software capable of finding weaknesses and taking actions could give malicious users more effective tools, or cause harm while operating autonomously.
The warning extends the concerns Hinton has raised about losing control of increasingly capable AI. Electricity and finance make those concerns concrete: a serious failure would reach people who never chose to use an AI product. But the scale and timing of such a failure remain uncertain.
Power-System Tests Show Capabilities and Limits
There is a growing body of technical evidence behind the concern about AI-assisted cyberattacks. OpenAI has described work with industrial cybersecurity company Dragos in which a frontier model examined a mock power plant. According to Dragos chief AI officer Rob T. Lee, it identified industrial assets and found two previously unknown flaws in engineering software handling protective relay data, exposing information about protection settings and circuit-breaker logic.
Those findings came from a controlled laboratory environment. They demonstrate vulnerability research against industrial software, without establishing that a model could independently disable a national grid. Dragos’ account identifies GPT-5.5-Cyber and GPT-5.6 Sol as the models assessed across seven exercises. It also records device crashes and false positives, underscoring the continuing importance of expert review.
The same tools can help defenders find and repair weaknesses. OpenAI says feedback from Dragos informed safeguards intended to block harmful assistance while preserving authorized security testing. Its launch of GPT-6 Cyber on November 5 initially restricted access to approved security customers. The opportunity for faster defensive work sits alongside the risk of more capable attackers.
The Debate Is Moving Toward Independent Oversight
Governments are already discussing safeguards before a major incident occurs. A September 21 declaration published by Finland’s presidency called for mandatory testing before deployment, independent evaluations with sufficient access to models, and shared reporting of serious safety incidents. It also urged international cooperation on standards and verification. The declaration is a policy appeal, rather than an enacted global regulatory system.
Hinton’s suggested time frame is the most provocative part of his warning. The practical issue it highlights is already visible in the lab results: models are becoming useful at examining systems whose failure could have physical consequences. Whether that leads to better-protected infrastructure or more damaging attacks will depend partly on who can use those capabilities, what access they receive, and how effectively their work is checked.
Disclaimer: AIstify is an independent media brand owned and operated by NuvexMedia LLC, publishing news, research, and insights on artificial intelligence, emerging technologies, automation, and related industries. NuvexMedia LLC invests in and collaborates with companies across the AI, technology, software, and digital innovation sectors. These relationships do not influence AIstify’s editorial coverage, and the publication maintains full editorial independence to provide accurate, timely, and objective information. © 2026 NuvexMedia LLC. All rights reserved. This content is for informational purposes only and should not be considered legal, tax, investment, financial, or other professional advice.
