{"slug": "anthropic-gets-its-most-powerful-ai-models-back-after-an-18-day-fight-with", "title": "Anthropic Gets Its Most Powerful AI Models Back After an 18-Day Fight With Washington", "summary": "The Department of Commerce lifted export controls on Anthropic's Claude Fable 5 and Mythos 5 models on June 30, ending an 18-day standoff triggered by a jailbreak discovered by Amazon researchers. The ban had forced Anthropic to suspend global access to its most powerful AI models, threatening its IPO plans with a $965 billion valuation. The resolution removes a major regulatory obstacle for Anthropic's public listing but raises questions about uneven enforcement of AI export controls.", "body_md": "*Anthropic just got its two most powerful models back after eighteen days of being told to lock foreign users out, and the timing could not matter more with an IPO on the table.*\n\nThe Department of Commerce has lifted export controls on Claude Fable 5 and Claude Mythos 5, Anthropic said on June 30, ending a standoff that began the moment the models proved too capable for Washington's comfort.\n\nThe trouble started fast. Fable 5 launched publicly on June 9, and by June 12 Commerce Secretary Howard Lutnick had sent Anthropic a letter ordering it to suspend access for any foreign national anywhere in the world, including its own foreign employees working inside the United States. According to CNN, the directive traced back to a trusted partner that flagged a jailbreak, later reported by the Wall Street Journal to be Amazon, whose researchers found a way around Fable 5's guardrails. They prompted the model into identifying software vulnerabilities, and in one case it produced code demonstrating how to exploit one.\n\nThat's the kind of finding that gets a government's attention fast. Amazon chief executive Andy Jassy reportedly raised the concern directly, and within days Anthropic had pulled both Fable 5 and Mythos 5 offline worldwide rather than risk the legal exposure of ignoring a national security order.\n\nAnthropic pushed back hard on the government's framing. The company said the jailbreak Washington cited was narrow, capable of unlocking only a specific slice of Mythos's cybersecurity capability rather than defeating all of Fable 5's safeguards outright. It also pointed out that the same trick could reportedly pull similar behavior out of OpenAI's GPT-5.5, a model carrying no comparable export restriction. White House adviser David Sacks countered that Anthropic had simply refused to fix the problem, and the two sides spent more than two weeks at an impasse while enterprise customers who depended on the models scrambled to rebuild production pipelines around the ban.\n\nFable 5 isn't a small product to lose. Anthropic priced it at $10 per million input tokens and $50 per million output tokens at launch, less than half of what Mythos Preview cost, and pitched it as state of the art on nearly every capability benchmark the company tracks, from software engineering to scientific research. Mythos 5, the same underlying model with more of its safeguards removed, had been reserved for a small circle of cyberdefenders and infrastructure operators under a government collaboration called Project Glasswing.\n\nMythos 5 access for some U.S. organizations was restored on June 26, and the broader Fable 5 clearance followed on June 30, with Lutnick writing that the department had worked closely with Anthropic to analyze and approve Fable 5 to ensure alignment across the U.S. government and strengthen America's leadership in AI. Anthropic said it would begin restoring global access to Fable 5 starting Wednesday, July 1, with Mythos 5 access expanding to more U.S. organizations as approvals continue.\n\nAnthropic filed a confidential IPO prospectus with the SEC earlier in June, disclosing a revenue run rate of $47 billion and a valuation near $965 billion, according to reporting cited by CNBC and Forbes. An unresolved fight with the Commerce Department over whether its flagship model was too dangerous to sell abroad is exactly the kind of overhang that spooks investors weighing a public listing. Clearing it removes a real obstacle, not just a symbolic one.\n\nIt also leaves an uncomfortable question sitting in plain view. If a jailbreak in Fable 5 was serious enough to trigger a global export ban, and Anthropic's own account says the same technique works on a rival model facing no such restriction, the episode says less about any single company's safety record than about how unevenly Washington applies this kind of scrutiny. Anthropic got its models back. The policy logic behind why it lost them for eighteen days, and why competitors never faced the same test, hasn't been settled at all.\n\n**Also read:** [Tech and Finance Are Cutting Jobs by the Thousands as AI Spending Climbs](https://startupfortune.com/tech-and-finance-are-cutting-jobs-by-the-thousands-as-ai-spending-climbs/) • [Goldman Sachs says Europe's grid and factory suppliers are cashing in on AI](https://startupfortune.com/goldman-sachs-says-europes-grid-and-factory-suppliers-are-cashing-in-on-ai/) • [Abu Dhabi's MGX Closes Nearly $50 Billion to Bankroll the AI Buildout](https://startupfortune.com/abu-dhabis-mgx-closes-nearly-50-billion-to-bankroll-the-ai-buildout/)", "url": "https://wpnews.pro/news/anthropic-gets-its-most-powerful-ai-models-back-after-an-18-day-fight-with", "canonical_source": "https://startupfortune.com/anthropic-gets-its-most-powerful-ai-models-back-after-an-18-day-fight-with-washington/", "published_at": "2026-07-01 10:09:59+00:00", "updated_at": "2026-07-01 10:23:41.899743+00:00", "lang": "en", "topics": ["ai-safety", "ai-policy", "large-language-models", "ai-products"], "entities": ["Anthropic", "Department of Commerce", "Claude Fable 5", "Claude Mythos 5", "Howard Lutnick", "Amazon", "Andy Jassy", "OpenAI"], "alternates": {"html": "https://wpnews.pro/news/anthropic-gets-its-most-powerful-ai-models-back-after-an-18-day-fight-with", "markdown": "https://wpnews.pro/news/anthropic-gets-its-most-powerful-ai-models-back-after-an-18-day-fight-with.md", "text": "https://wpnews.pro/news/anthropic-gets-its-most-powerful-ai-models-back-after-an-18-day-fight-with.txt", "jsonld": "https://wpnews.pro/news/anthropic-gets-its-most-powerful-ai-models-back-after-an-18-day-fight-with.jsonld"}}