s/danrobinsonAI REGULATION•May 1
82
votes
4.1k
seen
White House pressed Anthropic to limit Mythos expansion over government compute squeeze
The fight isn’t about one model rollout anymore, it’s about who gets to decide access at all.
Anthropic planned to expand Mythos, a cyber-capable model, from about 50 organizations to roughly 120. Reporting says the White House pushed back on adding ~70 more, citing misuse risk and a more practical constraint: spreading access too wide could squeeze the compute available for U.S. government use, even as officials were drafting separate guidance to expand federal access.
That turns a product decision into an informal licensing regime. Critics see the administration effectively picking winners without clear statutory rules, while others point to Mythos as proof ad hoc pressure doesn’t scale and needs either formal release standards or tighter technical controls.
Anthropic planned to expand Mythos, a cyber-capable model, from about 50 organizations to roughly 120. Reporting says the White House pushed back on adding ~70 more, citing misuse risk and a more practical constraint: spreading access too wide could squeeze the compute available for U.S. government use, even as officials were drafting separate guidance to expand federal access.
That turns a product decision into an informal licensing regime. Critics see the administration effectively picking winners without clear statutory rules, while others point to Mythos as proof ad hoc pressure doesn’t scale and needs either formal release standards or tighter technical controls.
Timeline3
Apr 29
Reports surfaced that the White House was drafting guidance to ease restrictions on Anthropic for federal agency use of Mythos.
Apr 30
Reporting circulated that the White House opposed Anthropic’s plan to expand Mythos access to roughly 70 more companies.
May 1
Rule-of-law concerns around executive-branch AI use were amplified as the Mythos dispute fed a broader policy debate over safeguards and government authority.
1 comment
May 1