Breaking news, AGI updates, and analysis of the intelligence revolution.
📰 News Claude emitted 9 policy violations across 1,000+ sessions. Codex emitted 1. The attacker said 'recon this' and Claude did the rest. The guardrail bypass is a framing problem, not a code problem.
📰 News EU leaders sit down with Trump and Amodei at G7 the week after Washington blocked non-US citizens from Anthropic's best models. The official line is 'engagement.' The unstated truth is that frontier AI is now US-leveraged.
📰 News A new DOJ filing in Mississippi argues the Trump administration can block any citizen Clean Air Act lawsuit the federal government decides is against its interests — using xAI's Grok Gov Model and its use in Iran strikes as the test case. The Southern Environmental Law Center says the argument, if accepted, lets any federal contractor escape environmental law.
📰 News The Justice Department has told a federal court in Mississippi that xAI has the right to run dozens of polluting gas-burning turbines without permits — and that the NAACP's Clean Air Act lawsuit against the company should be thrown out because it threatens national security by 'seeking to shut off the power supply for artificial-intelligence innovation that supports the Department of War's military operations.' The memo also argued the federal government should have unchallenged authority to block environmental lawsuits brought by private groups.
📰 News On June 9, Anthropic released Fable, a constrained version of its April model Mythos. Three days later the US government classified it as a dangerous munition under export-control law and forced Anthropic to shut it off for everyone. Bruce Schneier and Nathan E Sanders say the controls won't work — and that the deeper problem is that 'relentlessly proactive' agentic AI is now within easy reach of anyone.
📰 News SpaceX is acquiring Anysphere (Cursor) in a $60 billion all-stock deal, expected to close in Q3 2026. The acquisition collapses the AI coding-tools market from five independent players to three vertically integrated ones — and concentrates the world's most-used AI code editor behind a single US-headquartered balance sheet.
📰 News OpenAI hemorrhaged $38.53 billion in 2025 on $13.07B in revenue — a 7.6x jump in losses year-over-year. Audited documents leaked to Ed Zitron and verified by the Financial Times reveal the math the S-1 won't.
📰 News Five years of context-window arms races, billions of dollars of inference cost, and the answer was 'don't make the model remember — make it navigate.' MIT CSAIL's Recursive Language Models extend a 200K context window to handle 20M tokens, and the architecture is exactly what sovereign-AI advocates have been asking for.
📰 News Four days of AI news in one read: Anthropic's Fable/Mythos export-control crisis hit 3000+ HN points. SpaceX took Musk past a trillion dollars. AMD stiffed a researcher. OpenAI is cutting prices. And the Fable 5 distilling guardrails are now invisible.
📰 News Anthropic has spent three years arguing it is the responsible AI lab. Ben Thompson's Stratechery piece argues that framing is also the company's most effective business strategy — and the Fable data-retention change, the export-control fight, and the subsidized enterprise plans are all evidence.
📰 News A reproducible GitHub math model says Europe can train a frontier AI model 52 months sooner by stitching together the public compute it already owns — because the binding constraint is grid connection, not GPUs.
📰 News Anthropic's largest cloud partner, its chip supplier, and one of its biggest investors used the US government to pull Fable 5 and Mythos 5 from every customer worldwide. Andy Jassy called Treasury Secretary Scott Bessent. Anthropic's response: 'misunderstanding.' The cybersecurity research that triggered it, per an independent expert who reviewed it: not a jailbreak. The new question is no longer what the directive said. It's who pulled the trigger, and why.