Today's AI news is defined by a historic standoff between Anthropic and the Pentagon, as CEO Dario Amodei formally refused to strip safety guardrails from Claude less than 24 hours before a government deadline — with Google's own employees joining the pushback. Meanwhile, Nvidia posts a record $43 billion quarterly profit cementing its dominance of the AI infrastructure buildout, and Google launches Nano Banana 2, a faster 4K-capable AI image generation model rolling out across Gemini.

Anthropic vs the Pentagon

Anthropic CEO Dario Amodei formally refused Pentagon demands to strip safety restrictions from its Claude AI, with a Friday deadline set by Defence Secretary Pete Hegseth now in play. The standoff puts a $200 million contract and Anthropic's defence industry relationships at risk — and has already prompted Google DeepMind employees to write a letter urging their own leadership to resist similar pressure.

Anthropic refuses Pentagon's new terms, standing firm on lethal autonomous weapons and mass surveillance — The Verge
Less than 24 hours before the Pentagon's ultimatum deadline, Anthropic formally declined DoD demands for unrestricted access to Claude, refusing to waive policies against autonomous weapons and mass surveillance.

Anthropic says it 'cannot in good conscience' allow Pentagon to remove AI checks — The Guardian
Anthropic pushed back on Pete Hegseth's demand to remove safety limits from Claude, risking cancellation of a $200 million military contract and possible classification as a supply chain risk.

The Pentagon is making moves against Anthropic — The Verge
The Defense Department began contacting major contractors including Boeing and Lockheed Martin to assess their exposure to Anthropic products, in apparent preparation for a potential blacklisting.

Google Workers Seek 'Red Lines' on Military A.I., Echoing Anthropic — The New York Times
Google DeepMind employees sent a letter to chief scientist Jeff Dean urging him to resist Pentagon pressure for unrestricted military AI access, warning the dispute could set a dangerous industry precedent.

Singularity Soup Take: This is the most significant AI governance confrontation in years — if the Pentagon succeeds in forcing any AI company to abandon safety limits, it sets a precedent that could reshape how every AI developer worldwide writes their acceptable-use policies.

Nvidia's Record Quarter

Nvidia posted $43 billion in quarterly profit for Q4 fiscal 2026, smashing Wall Street expectations again — though markets responded with muted enthusiasm amid concerns about customer concentration and questions over long-term returns on AI infrastructure investment.

Nvidia's Quarterly Profit Hits $43 Billion on Strong A.I. Chip Sales — The New York Times
Nvidia's annual profit hit $120 billion, the company said, with Q4 results far exceeding analyst expectations as Big Tech's unabated AI spending continues to flow directly to the chip giant.

Nvidia's Huang says Pentagon–Anthropic rift is 'not the end of the world' — CNBC
Jensen Huang weighed in on the Anthropic-Pentagon standoff during Nvidia's earnings week, downplaying the dispute while highlighting Nvidia's $5 billion investment commitment in Anthropic.

Singularity Soup Take: A $120 billion annual profit from selling AI infrastructure is historically unprecedented — and markets' muted response suggests investors are now asking harder questions about when AI capital expenditure translates into productivity gains.

New Models: Google Goes Faster

Google launches Nano Banana 2 model with faster image generation — TechCrunch
Google's successor to its viral Nano Banana image model brings sub-second 4K synthesis and real-time web sourcing, and is now the default model across Gemini and Google AI Mode.

Google reveals Nano Banana 2 AI image model, coming to Gemini today — Ars Technica
Google's new image model immediately replaces previous versions with pro-level visual fidelity at flash speed, delivering advanced subject consistency and production-ready specs for developers.

Singularity Soup Take: Nano Banana 2's combination of real-time web sourcing and sub-second 4K output marks a genuine step change for professional AI image workflows — this is no longer just a consumer toy.

China's AI Moves

DeepSeek withholds latest AI model from US chipmakers including Nvidia — The Indian Express
DeepSeek broke with standard industry practice by denying Nvidia and AMD early access to its upcoming flagship model for hardware optimisation, instead giving Chinese chipmakers including Huawei a weeks-long head start.

'A huge vote of confidence': London set to host OpenAI's largest research hub outside US — IT Pro
OpenAI announced plans to make its London office its largest research hub outside San Francisco, intensifying competition with Google DeepMind for UK AI talent.

AI is starting to worry CEOs now — Axios
A growing number of CEOs are expressing concern about the financial and strategic implications of AI, as Microsoft stock drops 15% this year and uncertainty mounts around AI's actual business impact.

Microsoft Research Introduces CORPGEN to Manage Multi-Horizon Tasks for Autonomous AI Agents — MarkTechPost
Microsoft's CORPGEN framework uses hierarchical planning and memory to manage long-horizon tasks for autonomous agents, achieving a 3.5x improvement over baseline systems in completion rate.

Today's Pulse: 9 stories tracked across 10 sources — The Verge, The Guardian, The New York Times, CNBC, TechCrunch, Ars Technica, The Indian Express, IT Pro, Axios, MarkTechPost