Anthropic’s Claude models return online, while OpenAI introduces a heavily restricted rollout of its next-generation flagship model under tight federal watch.
Following negotiations, the restrictions on Claude models were lifted on 30 June 2026. Anthropic restored global access to Fable 5 via aggressive safety classifiers that step hazardous coding prompts back to older model engines. However, the Mythos 5 remains heavily restricted to government-vetted US critical infrastructure organizations. To prevent future blunt federal interventions, Anthropic, Amazon, Microsoft, and Google formed a coalition to build a standardized AI Jailbreak Severity Framework.
On 2 June 2026, President Trump signed Executive Order 14409 (“Promoting Advanced Artificial Intelligence Innovation and Security”). This directive established a national security framework classifying advanced systems meeting classified cybersecurity thresholds as “covered frontier models,” requesting private labs provide the government with exclusive pre-release testing access for up to 30 days.
The friction of this regime surfaced earlier on 12 June 2026, days after Anthropic unveiled its Claude Fable 5 and Mythos 5 models. Following an emergency alert regarding an advanced software vulnerability “jailbreak” exposed by Amazon researchers, the Commerce Department issued strict export-control directives, suspending access to foreign nationals—including Anthropic’s overseas staff. Lacking real-time nationality verification tools, Anthropic pulled both models offline globally for an 18-day shutdown.
Meanwhile, the policy shift altered rival rollouts. OpenAI delayed the broad public launch of its flagship GPT-5.6 Sol model, routing it instead into a limited preview for government-vetted partners. OpenAI stated it was cooperating short-term to establish a repeatable process but warned that “government access processes should not become the long-term default” by withholding vital defensive tools from public cyber defenders.
