Inside the Fable 5 Ban: How Independence Week’s Ugliest AI Fight Unfolded
With Fable 5 and Mythos 5 back online, the weekend’s reporting turned to how the fight actually happened. The trigger: researchers at Amazon discovered a “fix this code” jailbreak that could coax dangerous outputs from both models, and CEO Andy Jassy escalated it directly to Treasury Secretary Scott Bessent, Commerce Secretary Howard Lutnick, and National Cyber Director Harry Coker Jr. Lutnick then sent a letter to Anthropic CEO Dario Amodei imposing export controls, and the models went dark on June 12.
What followed was three weeks of crisis diplomacy: Anthropic flew engineers to Washington, NSA testing proceeded under the restricted Project Glasswing program, and Bessent and Lutnick ultimately helped defuse the standoff. The order lifted June 30 and Fable 5 returned worldwide July 1, only after Anthropic shipped a new safety classifier that blocks the reported jailbreak in more than 99% of cases. The lesson landing over the holiday: a frontier model is now close enough to critical infrastructure that a single jailbreak can put it on a cabinet secretary’s desk within days.
Anthropic Moves to Shut the Offshore Backdoor Chinese Firms Used to Reach Claude
According to a Financial Times report, Anthropic is stepping up efforts to detect and shut down the workarounds Chinese companies have used to access Claude despite its bans. The methods being targeted are creative: relay “transfer station” services that quietly route requests through overseas Claude accounts, foreign subsidiaries spun up to hold accounts, and access via cloud infrastructure including Microsoft Azure. Anthropic is reportedly monitoring signals like account time zones to spot the traffic.
The specifics name names: Ant Group reportedly gave employees corporate Claude accounts tied to a Singapore-based entity, while ByteDance reimbursed engineers for personal subscriptions accessed over VPNs. None of it appears to break US or Chinese law — it breaches Anthropic’s terms of service, which bar Chinese firms and entities under their control from using its models. The move extends earlier direct-access bans down to ownership structures and relay routes, a sign of how much enforcement work sits behind a simple line in a ToS.
Anthropic and OpenAI Took Roughly 43% of a Record $510B VC Half
The money story of the year keeps getting more lopsided. Per H1 2026 funding tallies making the rounds this weekend, global venture capital hit a record ~$510B in the first half of the year — and Anthropic and OpenAI together accounted for about 43% of it. Two companies, nearly half of all venture dollars deployed worldwide.
Read it next to the weekend’s other headlines and the picture sharpens: the same models drawing cabinet-level national-security scrutiny and ToS-level geopolitical enforcement are also the ones vacuuming up an unprecedented share of global capital. Concentration of capability and concentration of capital are, increasingly, the same story.
A Quiet Holiday Weekend That Was Really About Power
No launches, no new models — and yet this may be the most consequential weekend of the summer for understanding where Claude sits in the world. Three threads, one knot: a frontier model treated as national-security infrastructure (the Fable 5 post-mortem), a lab policing its own borders against a rival superpower (the China loophole crackdown), and an unprecedented concentration of capital flowing to the two firms at the center of it (43% of a record half).
The through-line is that Anthropic is no longer just shipping software; it’s operating something governments test, adversaries chase, and markets fund like scarce strategic supply. That’s a very different company than the one that published a chatbot. The product roadmap will keep making headlines — but weekends like this are where the real story of the next few years is being written: capability has become geopolitics, and the terms of service have become foreign policy.