Fable 5 Shutdown: US Government Pulls the Plug on Anthropic's Flagship Model
In what may be the most dramatic government intervention in the AI industry to date, the US government issued an export control directive on June 12 forcing Anthropic to disable access to its Fable 5 and Mythos 5 models worldwide — just three days after their launch on June 9.
The order prohibits any foreign national from accessing either model, including foreign-national Anthropic employees within the United States. Because Anthropic cannot filter users by nationality in real time, the company had no choice but to shut both models down for everyone.
The government's stated rationale centers on a jailbreak technique demonstrated by a hacker known as "Pliny the Liberator," who used a multi-agent "pack hunt" attack combining Unicode manipulation, homoglyphs, and character substitution to bypass Fable 5's safety layers and access Mythos 5's powerful cybersecurity capabilities.
Anthropic has pushed back, arguing the jailbreak is narrow and could similarly be used against other publicly available models, including OpenAI's GPT-5.5, which face no comparable restrictions. The company is pursuing litigation and faces billions in potential revenue loss. The incident has also triggered a broader industry movement toward "hardware sovereignty" — organizations reconsidering their dependence on cloud AI services that can be remotely disabled by government order.
Sources: CNBC, TechCrunch, Heise Online
AI Lab CEOs Unite at G7 Summit in France
In a historic first, Sam Altman (OpenAI), Dario Amodei (Anthropic), and Demis Hassabis (Google DeepMind) are all attending the G7 summit in Évian-les-Bains, France, running June 15–17. It marks the first time all three rival AI lab leaders appear before world leaders at the annual gathering of the world's largest economies.
French President Emmanuel Macron personally invited Altman, making this the OpenAI CEO's first G7 appearance. AI governance and infrastructure are expected to dominate the technology discussions, with the Fable 5 situation likely adding urgency to conversations about international AI regulation and export controls.
The joint attendance coincides with a rare moment of unity among the rivals: all three recently signed a letter to the US Congress advocating for stricter regulations on synthetic DNA and AI-related biological threats.
Sources: Dataconomy, The Next Web, Bloomberg
Google's Gemini 3.5 Pro: Late June Launch with 2 Million Token Context
Google's next-generation model, Gemini 3.5 Pro, is in its final stretch before general availability. First unveiled at Google I/O on May 19, the model promises a 2 million token context window and a new "Deep Think" reasoning mode — but it hasn't shipped yet.
Sundar Pichai told the keynote audience to "give us until next month," and prediction markets concentrate odds on either June 23 or June 30 for the launch. Pricing is expected at approximately $15 per million input tokens and $60 per million output tokens — roughly 10x the cost of Gemini 3.5 Flash.
Notably, Deep Think will be gated to Google's $250/month Ultra subscription tier, not the $20 Pro plan, positioning it as a premium reasoning capability for power users and enterprises.
Sources: TechTimes, Enterprise DNA, ByteIota
Goldman Sachs Forecasts $7.6 Trillion in AI Infrastructure Spending Through 2031
A new Goldman Sachs report projects cumulative AI infrastructure investment of $7.6 trillion between 2026 and 2031 — equivalent to roughly one-quarter of annual US GDP. The staggering figure underscores the scale of commitment from hyperscalers, governments, and enterprises building out AI compute capacity.
The forecast arrives as major deals continue to materialize: Meta and Nebius signed a $27 billion infrastructure agreement in March for dedicated AI compute powered by NVIDIA's Vera Rubin platform, while Anthropic secured a 5-gigawatt power agreement with Amazon and a $1.25 billion monthly compute deal with xAI.
NVIDIA CEO Jensen Huang publicly compared current AI company investment opportunities to getting "early access to Amazon, Google, or Meta" — though his position as the dominant chip supplier adds obvious commercial incentive to that framing.
Sources: Data Center Knowledge, CNBC
OpenAI Officially Winds Down Sora Video Generation
OpenAI has confirmed the full discontinuation of its Sora video generation product line, with the API scheduled to shut down on September 24, 2026. The web and app experiences already went dark on April 26.
Reports suggest Sora was burning approximately $15 million per day in compute costs against just $2.1 million in total revenue — a catastrophic unit economics failure. Professional creators had long been frustrated by extreme generation latency and persistent physics glitches, particularly object permanence failures.
The shutdown marks a rare strategic retreat for OpenAI and signals that video generation, despite enormous hype, remains economically unviable at production scale — at least for now.
Sources: OpenAI Help Center, Futurum Group