1
Anthropic, OpenAI and Google all opt in within a year
Likely
Resolves by Jun 2, 2027
Discussed by: Axios, NPR, CSO Online
Three labs have already praised the order publicly. The administration designed it around their existing voluntary work with the former AI Safety Institute, and the 30-day window matches their internal release cadences. If they sign agreements, the order has working content even without statutory backing.
2
A frontier model launches without federal pre-release access
Possible
Resolves by Jun 2, 2027
Discussed by: Tech Policy Press, Tech Oversight Project, Scientific American
xAI and Meta were not named as participating labs and their leaders lobbied to weaken the order. Either could ship a model that clears the NSA benchmark without going through the 30-day review, testing whether the administration is willing to apply public or commercial pressure.
3
Congress writes the framework into mandatory law
Uncertain
Resolves by Jan 20, 2028
Discussed by: Roll Call, CSO Online, Council on Foreign Relations
Senators on both parties have called for mandatory frontier AI rules since Biden's order was rescinded. A high-profile misuse, a Chinese AI scare or a presidential turnover could give Congress the political cover to convert the voluntary regime into a statute. The Council on Foreign Relations and Senate AI working groups have draft language already.
4
A successor administration rescinds the order
Possible
Resolves by Jun 2, 2029
Discussed by: Tech Oversight Project, Scientific American
Both Biden's and Trump's first-day AI moves were rescissions of the prior administration's order. A Democratic successor that wants stricter rules, or a libertarian-leaning successor that wants none, could repeat the pattern. The voluntary framework has no statutory floor to protect it.
5
Anthropic's Fable 5 and Mythos 5 remain suspended for months
Possible
Resolves by Q3 2026
Discussed by: Anthropic statement, TechRadar, Softonic, explainx.ai
Fable 5 and Mythos 5 were suspended June 12 under a Commerce Department export control directive. Anthropic says a single narrow, non-universal jailbreak does not justify pulling a model used by hundreds of millions. The government has given only verbal evidence and no public specifics. If the two sides cannot agree on what constitutes a sufficient fix, both models could stay offline for months — the first time that has happened to a deployed frontier model.
6
Export controls become a standing tool for frontier AI control
Possible
Resolves by Jan 1, 2027
Discussed by: Tech Policy Press, The Hill, Christian Science Monitor
The government used export control authority — not the voluntary framework — to pull Anthropic's models offline. If agencies repeat this with future launches, export controls would become a de facto mandatory review regime that bypasses the voluntary order entirely. The Anthropic case is the first known instance. A second use would establish the pattern.
