probability: 1%

rationale:

Short Time Window

The window for any new legislation to be drafted, passed, and enacted is extremely narrow. Current expectations point to Anthropic pricing its initial public offering around October 2026, with a projected median date of 26 October 2026 futuresearch.ai. Evaluating this tight expected timeline against the much longer multi-year horizon generally required to enact novel AI restrictions reduces the chance of such a law passing to just 1%. Even if the IPO timeline slips into early 2027 futuresearch.ai, there is virtually no time for an entirely novel category of binding legal restriction to pass through multiple legislative hurdles and enter into force in either the US or the EU.

Status Quo and Pending Legislation

As of late August 2026, no jurisdiction dictates which AI models developers may use for internal AI research 2 sources. Existing and pending laws heavily favor transparency, risk assessment, and downstream product regulation rather than internal use prohibitions. In the United States, enacted state laws—such as California's SB 53, New York's RAISE Act, and Illinois's SB 315—mandate frontier-AI frameworks, incident reporting, and safety protocols, explicitly avoiding restrictions on which models may be used for internal R&D 44 sources. Federal proposals like the FRONTIER Act (H.R. 9925) similarly focus on frameworks, audits, and catastrophic-risk reporting congress.gov.

Deregulatory Trends

Current federal and EU postures are actively moving away from imposing such mandates. In the US, Executive Order 14409 (June 2026) established a voluntary framework and explicitly disclaimed any mandatory licensing, preclearance, or permitting requirements for developing or releasing AI models whitehouse.gov. In Europe, the EU AI Act explicitly exempts AI models and systems developed and put into service solely for scientific R&D, as well as those in pre-market testing 2 sources. Furthermore, the EU's recent 2026 "digital omnibus" amendments delayed high-risk obligations rather than introducing new prohibitions ztekcyber.com.

Key Uncertainties and Tail Risks

Given the exceptionally low base rate for rapid, novel legislative action, the remaining 1% residual probability is driven almost entirely by tail risks and edge cases. Even under a rapid capability acceleration scenario that might otherwise prompt regulatory panic , an unprecedented AI or biosecurity incident triggering emergency legislation would still have to overcome the current US administration's stated no-licensing policy whitehouse.gov. Another edge case involves resolution ambiguity—such as the June 2026 US export controls that restricted certain foreign nationals' access to frontier models inside US labs ai-frontiers.org. While this functions as a national security access control rather than a general restriction on using AI for AI research, it highlights how quickly executive actions can be implemented. Finally, a significant delay in Anthropic's IPO into late 2027 would lengthen the resolution window, marginally increasing the feasibility of a new law taking effect.