Companies are reining in runaway AI spending, the EU stakes its claim to tech sovereignty, and fresh data shows which jobs AI is really changing.
The US government forced Anthropic to switch off two frontier models worldwide, turning export law into an AI kill switch. Plus DeepMind's UK planning AI, Grok on Bedrock, and PwC's split labour market.
Engineers at Amazon and Google mock their mandated AI as 'slop' even as adoption hits 88% and measured impact stays near 7%. Plus Fable 5, DiffusionGemma and the EU's tech sovereignty push.
A labour economist takes apart the AI jobs-apocalypse narrative this week, arguing the displacement story can't be found in the numbers, even as OpenAI, Microsoft and Mistral push agents into every office role.
Frontline workers distrust the AI dropped on them, and a Stanford study shows the hiring algorithms screening them carry measured racial bias, while Opus 4.8, Mistral's physics-AI buy, and DeepSeek's 75% price cut set the backdrop.
Anthropic overtakes OpenAI in paid business adoption, OpenAI spins out a $14B deployment arm, Bloomberg confirms heavy job losses in AI-exposed roles. Where the work actually happens is now the contested ground.
OpenAI shipped GPT-5.5 and Workspace Agents in the same week, Anthropic released Claude Design and Opus 4.7, Meta trained AI on 8,000 employees' keystrokes, and SpaceX took a $60B option on Cursor. Agents are no longer a demo.
Anthropic ships Claude Design and Opus 4.7, DeepMind releases Gemini Robotics-ER 1.6, and Stanford's AI Index puts organisational adoption at 88 percent. The frontier race has slowed and the integration race has started.