Anthropic got Fable 5 back after a ten-day suspension, but it came back different: nationality-based access controls, stricter safety classifiers, mandatory data retention. And today, June 22, the free trial window expires. Meanwhile, Norway has become the first country to nearly ban generative AI in elementary schools, and OpenAI Codex has hit 5 million weekly users — partly fueled by those who sought alternatives during the Fable 5 outage. Three stories about the same phenomenon: AI is entering its accountability phase, and markets, governments, and users are beginning to react.
Anthropic's most powerful model was restored on June 18, six days after a US government export control order forced its global shutdown. But the Fable 5 that came back is not the same one that launched on June 9. Anthropic implemented three changes that mark a before and after in how frontier models are deployed.
First, stricter safety classifiers: the restored model routes more cybersecurity, biology, chemistry, and distillation queries to Opus 4.8. The percentage of sessions running entirely on Fable 5 has dropped versus launch. Second, nationality-based access controls: for the first time, a commercial AI API implements government-mandated nationality verification. Third, mandatory data retention: the updated privacy policy (effective July 8) includes government-issued ID and biometric data collection for certain access tiers.
And here's the detail that stings most for developers: today, June 22, the Fable 5 free window expires. Starting tomorrow, using the model costs $10 per million input tokens and $50 per million output. Anthropic has not extended the free period despite the model being offline for 6 of the promised 14 days. The Anthropic community manager on X summed it up well: "Fable 5 came back, but it came back changed."
Norway has announced a near-total ban on generative AI tools for elementary school pupils — ages 6 to 13 — effective at the end of August. It is the first country to implement such a broad and direct restriction, setting a precedent that other European governments are already watching closely.
The measure distinguishes by age group: a general ban for first through seventh grade (ages 6-13), and significant restrictions for lower secondary school (up to age 15), where AI use will require explicit teacher authorization and be limited to specific pedagogical contexts. The Norwegian government argues that generative AI interferes with the development of foundational skills — writing, critical thinking, unassisted problem-solving — and that there is insufficient evidence about its long-term effects on children's learning.
Norway's decision raises a question more and more governments are asking themselves: are we letting AI into classrooms without really understanding what it means for children's cognitive development? While the UK is pushing to integrate AI into the curriculum and Singapore actively encourages it, Norway is pulling the emergency brake. And it probably won't be the last.
OpenAI has announced that Codex, its async coding agent, has reached 5 million weekly active users. The figure matters not just for the volume, but for the context: a significant portion of that growth happened during the six days Fable 5 was offline, when companies that depended on it migrated to Codex as an alternative.
Codex also launched Record and Replay for macOS, a feature that lets Business users record a workflow once and turn it into a reusable skill for Codex, Computer Use, browser actions, and plugins. It's a clear strategic move: OpenAI is building the developer agent platform, not just a chat that can write code.
The timing is symbolic. While Anthropic deals with government restrictions on its flagship model, OpenAI capitalizes on the gap and turns a competitor's vulnerability into product traction. The coding agents war has just begun, and Codex is making the most of it.
Three stories from seemingly different worlds, but they all tell the same story: artificial intelligence is leaving behind its limitless expansion phase and entering a maturation stage where checks and balances appear. Fable 5 returns with controls because a government decided national security trumps innovation. Norway bans AI in elementary schools because another government decided children's cognitive development cannot be put at risk without evidence. And OpenAI Codex grows because the market reshuffles when a competitor stumbles.
The lesson is clear: the "move fast and break things" era in AI is closing. We are entering a phase where regulation, accountability, and ecosystem resilience matter as much as raw model capability. And that, however uncomfortable for the growth-at-all-costs purists, is a sign of health.
Baseten doesn't train models — it runs them. Its platform lets any company deploy open-source and proprietary models with predictable latency, without having to manage GPUs themsel...
Read article →A $13 billion capital tsunami for a single startup. An aerospace giant buying a code copilot for $60 billion. And the birth of India's first AI unicorn, which is less a celebration...
Read article →88% of global AI investment stays in the United States. Meanwhile, India produces its first pure AI unicorn with Sarvam, and infrastructure startups raise $1.8 billion in 48 hours....
Read article →