You Can Argue With Rules. Not With Fog.

Microsoft's president just said the quiet part out loud: Washington is regulating frontier AI without rules anyone can read — and GPT-5.6 went public anyway on the same July 9 morning.

What's new: At Geneva's AI for Good Global Summit, Brad Smith told Fortune and The Next Web that U.S. AI policy has crossed a line.

• "What we really have right now is regulation without transparent or complete rules," Smith said. "Without rules, businesses can't plan."
• The same day, OpenAI launched GPT-5.6 Sol, Terra, and Luna broadly — ending a 12-day preview limited to roughly 20 government-vetted partners, per TechTimes and Channel News Asia.
• A White House official told AFP there was no administration "green light, approval or clearance" — even after Commerce's CAISI completed its review.

The one blunt tool: Smith defended Washington's instinct to act on cyber risk. He attacked the instrument.

• Commerce used export-control law in June to force Anthropic offline worldwide over Fable 5 and Mythos 5 — the first time Washington treated a live API model like a chip crossing a border.
• Weeks later, the White House pressed OpenAI to gate GPT-5.6 behind a vetted customer list — a different agency, different lever, same fog.
• "The government got information that there was an urgent cybersecurity risk. When it gets that information, I think it's right to act," Smith said. "But it found it had only one regulatory tool: an export control."
• "The government doesn't have the tools it needs," he added. "Common sense says don't be heavy-handed, but have enough of a touch to do what's needed."

Two cloud chiefs, opposite reads: At the same summit, the industry's biggest buyers split on whether opacity is a feature or a bug.

• Smith said allies misread the Fable shutdown as a foreign blockade. Washington, he argued, wanted the model pulled "at home and abroad."
• Marc Benioff took the other side — he felt "good" about the block and said Europe had misread a security call as hostility, per TNW.
• Canadian Prime Minister Mark Carney had already framed the stakes in June: "The situation we're in collectively right now with Mythos and Fable is something that can happen with overreliance on certain models," per the Associated Press.

Why it matters now: GPT-5.6 is the first frontier family to complete Trump's June 2 voluntary pre-release framework — legally optional, commercially coercive.

• President Trump's Executive Order 14409 disclaims "mandatory governmental licensing, preclearance, or permitting."
• In practice, Commerce Secretary Howard Lutnick sought assurances from Sam Altman that federal agencies had finished review before any wider rollout, per TechTimes.
• OpenAI still warned it does not want this process to become "the long-term default," per CNA — weeks before both labs filed confidential IPO paperwork targeting ~$1 trillion valuations.
• Smith's bottom line for Azure's global customers: "People will not buy what we sell unless they're confident of certainty of supply."

What to watch:

• August 1: The EO's deadline for NSA's classified frontier benchmark and Commerce's voluntary framework — or another month of improvised gates.
• Five-lab talks reported by the Financial Times toward a shared jailbreak severity score — the first attempt to replace phone calls with a published standard.
• Grok 4.5 shipping the same week from SpaceXAI without an announced federal review — a live test of whether the gate applies to everyone or just the IPO-bound.

Convina's view: Silicon Valley spent two years calling for light touch. Now its most global salesman is begging for a rulebook — because improvised kill switches are worse for procurement than statutes you can litigate. Washington built a licensing regime without writing one down. That is not deregulation. It is fog with export-control teeth — and the first company that ships a frontier model without asking permission will force everyone to admit the gate was never law to begin with.