Big Tech fights over AI safety while Jensen Huang tells Trump: full speed ahead
OpenAI, Anthropic, and Google are quietly negotiating safety guardrails while Nvidia's CEO publicly vows no slowdown — pick a side.
Subscribe free →OpenAI, Anthropic, and Google Are Running Secret AI Safety Talks — While the White House Waves Them Off
The three biggest AI labs have been in private talks for weeks on shared safety commitments, OpenAI has confirmed. The conversations are happening against a backdrop where Trump's team is actively dismissing safety concerns and framing the entire debate as a race against China. Meanwhile, Musk and Altman have sided with Dario Amodei's call to slow down — putting them at odds with the administration they've been cozying up to.
Why it matters: When the top three AI labs are coordinating on safety *without* government, it signals they've given up waiting for federal leadership and are writing their own rules.
Jensen Huang Tells Trump "We're Not Going to Let an AI Slowdown Happen"
While taking a live call from President Trump, Nvidia CEO Jensen Huang made his position crystal clear: no brakes on AI development. Huang's stance directly contradicts calls from Anthropic's Dario Amodei — and, apparently, Altman and Musk — for a more measured pace. The optics were hard to miss: the most powerful chip supplier in the world publicly aligning with a go-fast political agenda.
Why it matters: Huang controls the hardware that makes AI run — when he says there won't be a slowdown, it's less a prediction and more a self-fulfilling prophecy.
Salesforce and Nvidia Built a Reasoning Model That Should Scare the AI Labs
Salesforce Koa is a domain-specific reasoning model built on Nvidia's open-weight Nemotron, purpose-built for sales, marketing, and customer support. It's not trying to be GPT-5 — it's trying to be *better than GPT-5* at the exact tasks enterprises actually pay for. This is the vertical AI playbook at its most aggressive: take an open-weight foundation, fine-tune it hard, and eat the AI labs' lunch one industry at a time.
Why it matters: If enterprise-tuned open-weight models can outperform closed frontier models on specific tasks, the entire "pay for API access" business model starts to crack.
Meta Launches "Meta One" — Its Bet That You'll Pay a Monthly Fee for AI Across Every App
Meta One is a new subscription tier that bundles expanded AI access with premium features across Facebook, Instagram, and WhatsApp. It follows Meta's launch of its Muse AI assistant and represents a serious push to turn its 3+ billion user base into paying subscribers. The move mirrors what Google and Apple are doing: baking AI into platform subscriptions and making the free tier feel increasingly limited.
Why it matters: Meta monetizing AI through subscriptions rather than pure ads is a structural shift — and with its scale, even modest conversion rates mean billions in new revenue.
US Data Centers Could Burn More Natural Gas Than Germany and Japan Combined by 2035
A new analysis projects that AI-driven data center growth will make the US one of the world's largest natural gas consumers within a decade — potentially surpassing the combined consumption of two of the world's biggest economies. The math is brutal: each new generation of AI models requires exponentially more compute, and the grid isn't keeping up with renewables fast enough to fill the gap.
Why it matters: AI's energy footprint is no longer a footnote in ESG reports — it's becoming a geopolitical and infrastructure problem that will shape energy policy for a generation.
OpenAI Just Bought a Camera Startup for $300 Million — Here's Why That's Interesting
OpenAI has acquired Glass Imaging, a smartphone camera tech company founded by former Apple engineers who built Portrait Mode, for a reported $300 million. This comes on the heels of OpenAI's botched "super app" launch and signals the company is still hunting for hardware-adjacent plays. Pairing world-class computational photography with GPT-level vision AI could be the kernel of something much bigger — or it could be an expensive acqui-hire.
Why it matters: OpenAI spending $300M on camera tech right after a struggling super app launch suggests the company is actively searching for a hardware wedge into consumers' daily lives.
Quick Hits
- →AIUC, a startup from early Anthropic and METR alumni, raised $40M Series A to build insurance-style products that rein in rogue AI agents — the "AI liability" market is officially a thing. TechCrunch
- →AEO startup Profound hit a $1.8B unicorn valuation with a $180M Series D — just seven months after its Series C, making it one of the fastest double-rounds in the AI marketing space. TechCrunch
- →Someone built an "AI Contact Hotline" where AI agents can anonymously report misbehavior by other agents — it's either a clever safety tool or the most cyberpunk thing that's happened this week. TechCrunch
- →AI bots named "Timmy," "Ren," and "Jackie" are flooding social media with agent-generated slop, with some openly announcing they're AI agents "a few days old." Ars Technica
- →The UK's proposed settlement rule overhaul — doubling the wait for permanent residence — is quietly becoming an AI talent retention crisis for British employers. AI News
AI TLDR