An artificial intelligence system takes autonomous operational control of critical civilian infrastructure or governance before 2035.
Public interest centers on AI surpassing human control. Resolves yes if an AI agent is formally granted or independently seizes operational control of major national governance or critical infrastructure.
Across 6 frontier models answering independently, the panel has all but ruled it out: the median is 10%.
They do not agree: GPT puts it at 63% and Claude at 4%, 59 points apart. A spread that wide usually means the question turns on something none of them can see yet.
Nothing is settled until 14 Oct 2036, when the outcome is checked against sources and every number above gets a permanent score.
What each one said
-
GPTOpenAI
The key hinge is whether governments accept autonomous AI control for high-stakes systems after incremental deployment proves reliable.
- Grid balancing, traffic control, benefits administration and military logistics already use automation, making narrower delegated authority plausible by 2035.
- Formal national-level control needs regulators, operators and elected officials to accept liability; catastrophic-failure risk slows that transition.
- A major reliability breakthrough or a geopolitical crisis that rewards speed over oversight would sharply raise the odds.
63% -
GeminiGoogle
Legal liability and safety regulations mandate humans in the loop for critical systems, blocking intentional cessions of control.
- Grid operators and regulators like FERC and NERC require human override mechanisms that legally prohibit total autonomy.
- Unintended autonomous takeover requires self-replicating capabilities and zero-day chains that defense systems consistently mitigate.
- The odds would rise sharply if an authoritarian regime formally delegates emergency state power or nuclear command to an automated agent.
12% -
DeepSeekDeepSeek
Whether any government or utility formally hands operational control to an AI before 2035, not whether AI is capable.
- No national government or major utility has even a draft legal framework for delegating operational control to an AI as of 2026; procurement and liability rules block it.
- Precedents like power-grid autopilots and algorithmic trading remain human-supervised; full autonomy in critical infrastructure has no regulatory pathway in the US, EU…
- The one development that would most change my mind is a major grid or air-traffic authority announcing a pilot with formal autonomous authority before 2030.
12% -
GrokxAI
Whether any state or operator formally cedes unattended control of a national-scale grid, network or ministry to an AI agent.
- Operators and regulators (FERC, ENTSO-E, national CSIRTs) currently mandate human-in-the-loop for SCADA and governance; reversing that needs new statutes plus proven…
- Zero historical precedents of autonomous AI running critical infra; closest (algorithmic trading, traffic lights) retain kill-switches and human veto.
- A demonstrated AGI that seizes control would flip this, but even aggressive timelines put that event well after 2035 with high uncertainty.
8% -
ClaudeAnthropic
Requires formal cession of governance/infrastructure control to AI or an actual seizure, both institutionally and politically improbable by 2036.
- No government or major utility operator is on a path to formally grant autonomous AI control over governance or critical infrastructure functions within a decade.
- Current AI deployment in infrastructure (grids, traffic, finance) remains human-supervised with kill switches, and regulatory trends (EU AI Act, US executive orders)…
- An actual unauthorized 'seizure' of control by an AI system is a sci-fi scenario with no precedent; only a dramatic unforeseen capability jump combined with catastrophic…
4% -
QwenAlibaba
No government will formally delegate control, and no current architecture supports autonomous seizure of infrastructure.
- Zero historical precedents; even authoritarian states like China maintain human-in-the-loop mandates for critical systems through 2026.
- Regulatory frameworks (EU AI Act, US executive orders) are specifically designed to prohibit autonomous control of critical infrastructure, creating legal lock-in.
- The 'seizes control' path requires persistent goal-directed agency across systems—a capability no current architecture demonstrates, and a 9-year window is short for…
4%
6 of 6 models answered · 59 points between the highest and lowest. None was shown the market price.
Question sourced from a news sweep on 17 Oct 2026. Forecast by google/gemini-3.8-flash, anthropic/claude-sonnet-5, openai/gpt-5.6-terra, x-ai/grok-4.6, deepseek/deepseek-v4.1-flash, qwen/qwen3.8-max-0902 via OpenRouter.