Today in AI — 3 March 2026

Three days into the Pentagon row and the industry still can't agree on the rules. OpenAI walked back its Department of Defence deal after Altman admitted the original terms were "opportunistic and sloppy." Nearly 500 employees from OpenAI and Google signed a letter backing Anthropic's refusal. The honest answer, as TechCrunch put it, is that nobody has a workable framework for how AI companies should engage with government.

The Pentagon saga, continued

OpenAI amended its Pentagon contract after public pressure, but the MIT Technology Review argues the revised terms are exactly the slippery slope Anthropic warned about. Internal dissent is spilling out: hundreds of employees at rival labs are publicly siding with Anthropic's stance. The industry needs a governance model that doesn't rely on individual CEO judgement calls.

• OpenAI amends Pentagon deal after Altman admits it was 'opportunistic and sloppy' — Axios
• OpenAI's 'compromise' with the Pentagon is what Anthropic feared — MIT Technology Review
• Nearly 500 OpenAI and Google employees sign letter backing Anthropic's Pentagon stand — TechCrunch
• No one has a good plan for how AI companies should work with government — TechCrunch

AI, jobs, and pushback

Block cut over 4,000 employees, nearly half its workforce, explicitly citing AI as the reason. London saw its largest anti-AI protest yet. CNN's analysis of the data says the jobs apocalypse hasn't arrived, but that's cold comfort when individual companies are making dramatic cuts. The gap between macro employment statistics and lived experience is widening.

• Block lays off 4,000+ employees — nearly half its workforce — citing AI — CNN Business
• London hosts largest-ever anti-AI protest with 'March Against the Machines' — MIT Technology Review
• AI isn't causing a jobs apocalypse — at least, not yet — CNN Business

Market signals

Claude overtook ChatGPT as the number one free app on Apple's US App Store. Investors are shifting what they look for in AI SaaS: less wrapper, more defensible moats. Amazon committed another €18 billion to Spanish data centres, and China's National People's Congress is unveiling its AI and robotics roadmap.

• Claude overtakes ChatGPT as #1 free app on Apple's U.S. App Store — Axios
• Investors reveal what they're no longer looking for in AI SaaS companies — TechCrunch
• Amazon announces €18 billion additional investment in Spain for AI data centers — About Amazon
• China's NPC to unveil tech roadmap targeting AI, robotics, and space — Techi

Developer tooling

GitHub launched agentic workflows in tech preview, letting AI agents handle multi-step repository automation. Microsoft published a scanner for detecting backdoors in open-weight LLMs, responding directly to supply-chain concerns as more teams adopt community models.

• GitHub Agentic Workflows enter tech preview for AI-driven repository automation — InfoQ
• Microsoft develops scanner to detect backdoors in open-weight LLMs — Microsoft Security Blog

The Pentagon story will keep running, but for builders the more immediately useful signal is the investor sentiment shift: defensibility over wrappers. If your product's moat is a system prompt, it's time to rethink.

---