⏎ Words Summary from News
**The US government has partially reversed a ban on Anthropic's powerful Mythos 5 AI model, allowing over 100 trusted US organizations—including Fortune 500 companies—to access it for critical infrastructure defense.** The original June 12 export control order abruptly disabled Mythos 5 and Fable 5 over fears that adversaries like China or Russia could weaponize the models for sophisticated cyberattacks. Anthropic confirmed the redeployment on Friday, stating it is working to expand access and eventually restore Fable 5 for general use. The decision follows a similar delay of OpenAI's GPT-5.6, which is now limited to vetted partners.</p><p class="summary-lead">**Critics argue the government's opaque selection process for approved companies concentrates excessive power in Washington and undermines transparency.** Free speech advocate John Coleman noted that no one knows how companies are chosen or why others are excluded, raising rule-of-law concerns. OpenAI CEO Sam Altman echoed this, stating he dislikes the idea of the government picking customers, even if safety testing is warranted. The Commerce Department cited “significant progress” in addressing risks but did not specify what safeguards were adopted.</p><p class="summary-lead">**The underlying tension pits national security against innovation, with experts warning that prolonged restrictions could allow China to close the AI gap.** Mythos models, if misused, could dramatically accelerate cyberattacks on sectors like banking that rely on aging, interconnected systems. The government is moving toward allowing Fable 5's release, though no timeline is set. Analyst Kate Koren called the current approach a practical interim step but stressed that without a clear system for wide release, US companies risk falling behind global competitors.</p><p class="summary-lead">**What to watch next:** Whether the government formalizes a transparent approval framework for frontier AI models, and how Anthropic and OpenAI navigate the balance between public availability and national security demands.
Key Takeaways
- The US partially lifted its ban on Anthropic's Mythos 5, granting access to over 100 trusted organizations for critical infrastructure defense.
- Critics decry the lack of transparency in the government's selection process, raising concerns about executive overreach and rule of law.
- Prolonged restrictions on advanced AI models risk ceding competitive ground to China, according to analysts.
- Both Anthropic and OpenAI face ongoing government pressure, with OpenAI delaying GPT-5.6's public launch under similar terms.
Insights & Analysis
- The selective access model creates a two-tier AI ecosystem where only government-vetted entities benefit from cutting-edge capabilities, potentially stifling broader innovation.
- This episode signals a permanent shift in US AI policy: frontier models will likely never see unfettered public release, forcing companies to design for compliance from the outset.