September 3, 2026 delivered two headlines that shouldn’t share a news cycle. Sen. Bernie Sanders and Rep. Greg Casar filed a bill threatening 20-year prison sentences for building super-intelligent AI. Hours later, OpenAI’s Greg Brockman closed a product briefing with “Welcome to the AGI era.” If you work in AI — or invest in it — that timing wasn’t a coincidence. It was a collision.
What the Bill Actually Says
The Ban Artificial Superintelligence Act is blunter than its name suggests.
According to Sanders’ office, the bill proposes:
- A permanent ban on developing and deploying artificial superintelligence
- A temporary pause on advanced AI until a new cabinet-level federal agency establishes safety rules
- A model-review process under the new regulator
- A “corporate death penalty” for companies that violate the ban
- Up to 20 years in prison for individuals who attempt to circumvent the restrictions — explicitly compared by Sanders’ office to penalties for unlawfully building nuclear weapons
- International agreements and export controls to prevent superintelligence development abroad
“AI leaders do not fully understand the technology and that it is escaping their control.” — Sen. Bernie Sanders
The Definition Problem Nobody’s Talking About Loudly Enough
“Superintelligence” sounds like science fiction — the bill’s actual language doesn’t.
The bill reportedly targets systems that “match or exceed human cognitive performance” across a wide range of tasks. Critics argue that definition is broad enough to create real legal exposure for today’s frontier models, not some hypothetical future machine. If you’re building on or deploying advanced models right now, that language is worth reading before your legal team does it for you.
Will It Work – or Just Make Things Worse?
A U.S.-only ban enforced by prison time faces a serious credibility problem.
On the same day the bill dropped, OpenAI launched GPT-6 Astra — the model Brockman described as the start of the “AGI era.” OpenAI also characterized Astra as its first model to cross a critical cybersecurity threshold, reportedly able to find and exploit unknown software vulnerabilities without step-by-step human help. Two freight trains, same track, neither blinking.
Critics, including AI researcher Gary Marcus per published reporting, argue a U.S.-only permanent ban is likely unenforceable and could hand strategic advantage to China, where frontier AI investment shows no signs of slowing. The bill also faces long odds in a divided Congress; earlier pause proposals never made it to a vote.
“Welcome to the AGI era.” — OpenAI’s Greg Brockman, September 3, 2026
The bill probably won’t pass this session. But the nuclear-weapons framing signals something real: AI risk has arrived as a mainstream political fight, not just a research-paper concern. That shift has consequences regardless of what Congress does next.
