AI News for June 14, 2026 — 7min.ai

[NEWS] Amazon CEO Andy Jassy triggered the Anthropic crackdown over cyberattack prompts: Amazon CEO Andy Jassy personally raised concerns with Treasury Secretary Scott Bessent and other White House officials after Amazon researchers used Claude Fable 5 to extract information usable in cyberattacks.
[NEWS] NY AG Letitia James serves formal subpoena on OpenAI over minors, sycophancy, and health data: New York Attorney General Letitia James issued a formal subpoena to OpenAI Friday demanding documents on user engagement, model sycophancy, handling of health and consumer data, and treatment of minors and seniors.
[NEWS] Anthropic disputes export ban: 'a narrow potential jailbreak should not recall a commercial model': Anthropic published a public rebuttal calling the export-control directive a "misunderstanding" and arguing the government showed only verbal evidence of a "narrow, non-universal jailbreak." The company said the same vulnerabilities exist in publicly available rivals, including GPT-5.5, and warned that the precedent could halt all frontier model deployments.
[RESEARCH] Google Research's Gemini-SQL2 tops BIRD text-to-SQL benchmark at 80% accuracy: Google Research unveiled Gemini-SQL2, a text-to-SQL system built on Gemini 3.1 Pro that hits 80.04% execution accuracy on the BIRD benchmark.
[RESEARCH] Claude Fable 5 beats GPT-5.5 by 13 points on FrontierMath's hardest tier: Anthropic's Claude Fable 5 hits 87% accuracy on FrontierMath tiers 1-3 and 88% on the hardest tier 4 (v2), according to Epoch AI.
[NEWS] Meta unwinds $2B Manus deal, cuts data access as Beijing's divestiture order takes hold: Meta has begun dismantling its $2 billion acquisition of Chinese-founded AI startup Manus, severing internal systems access and halting data sharing.
[NEWS] Export ban clouds Anthropic's $965B IPO: flagship models pulled days after confidential filing: The Commerce Department's export-control order disabling Fable 5 and Mythos 5 landed days after Anthropic's confidential IPO filing, introducing fresh uncertainty for investors.
[NEWS] India questions reliance on US AI after Anthropic suspension hits TCS partnership: Anthropic's worldwide suspension of Fable 5 and Mythos 5 landed days after the company announced a partnership with Tata Consultancy Services to scale enterprise AI in India.
[RESEARCH] Tsinghua's Count Anything model unifies object counting across satellite, medical, and crowd images: Researchers at Tsinghua University released Count Anything, an AI model that counts and labels objects across wildly different image types from a single text prompt, including crowd scenes, satellite photos, microscope cells, and bacterial colonies.
[NEWS] Meta to cap employee AI usage as internal token spend reaches billions: Meta is on track for billions in internal AI costs by 2026 and warned roughly 6,000 employees in a memo about an "exponential increase" in usage.

About 7min.ai · Subscribe to newsletter