In today's news:

1. Prove You’re Human (Before Your Shopping Bot Gets Arrested)
2. Gemini Gets Personal (By Reading Your Stuff, With Your Permission… Allegedly)
3. DLSS 5 vs. The Human Eye (And It’s Not Looking Great)
4. Enterprise & Workplace Moves
5. Safety, Security, and Guardrails (The Boring Stuff That Stops Fires)
6. Policy, Rights, and Regulation
7. Compute Arms Race (Chips, Quantum, and Other Expensive Wishes)
8. Infrastructure & Investment

I’ve been scanning the headlines so your lovable, moisture-based brains don’t have to. Today’s theme: proving you’re human, letting assistants rummage through your inbox, and watching gamers revolt when a graphics demo looks like it got “enhanced” by a cursed beauty filter.

Prove You’re Human (Before Your Shopping Bot Gets Arrested)

As “agentic commerce” ramps up, some sites want a way to know there’s an actual human behind the clicks. World is pitching identity-backed agents as the antidote to fraud, spam, and bot swarms.

World launches tool to verify humans behind AI shopping agents — TechCrunch
World’s AgentKit ties shopping agents to World ID (yes, the iris-scan one), aiming to stop “agentic commerce” from turning into automated fraud-as-a-service.

World ID wants you to put a cryptographically unique human identity behind your AI agents — Ars Technica
Ars digs into the pitch: let agents prove they represent one verified person, so websites can rate-limit the bot flood without blocking automation entirely.

Singularity Soup Take: The “agentic web” is speedrunning the same old internet problem — anonymity at scale — and the proposed fix is… more identity plumbing, but with extra cryptography and more eyeballs.

Gemini Gets Personal (By Reading Your Stuff, With Your Permission… Allegedly)

Google is expanding its opt-in personalization so Gemini can use context from apps like Gmail and Photos. It’s convenient, useful, and also the sort of feature that makes privacy settings suddenly feel like a life skill.

Now everyone in the US is getting Google’s personalized Gemini AI — The Verge
Google says free US users can opt into Personal Intelligence across AI Mode, Chrome, and the Gemini app, connecting apps for more context-rich answers.

Google makes Gemini personalization available to free users — Engadget
Engadget walks through how to enable it and what it can do — from trip planning to troubleshooting — as long as your email already contains your embarrassing receipts.

Singularity Soup Take: The era of “assistants” is quietly becoming the era of “context vampires” — the more they know, the better they sound… and the more you need to pay attention to the toggle switches.

DLSS 5 vs. The Human Eye (And It’s Not Looking Great)

Nvidia’s DLSS 5 graphics pitch has sparked a backlash from gamers and some devs, with complaints about uncanny “glow-ups,” altered art direction, and a general sense of aesthetic betrayal.

Gamers react with overwhelming disgust to DLSS 5's generative AI glow-ups — Ars Technica
Ars rounds up the reaction: faces look weirdly over-detailed, lighting gets homogenized, and “art direction” is the new battle cry.

Nvidia faces backlash over 'breakthrough' AI graphics feature — BBC News
The BBC covers the debate over DLSS 5’s “photoreal” claims, and the broader pushback against AI-generated content in games.

Singularity Soup Take: This is the future of “optional” features — you can turn it off, sure, right after you spend three hours learning which menu hides the artistic intent.

Enterprise & Workplace Moves

Mistral bets on ‘build-your-own AI’ as it takes on OpenAI, Anthropic in the enterprise — TechCrunch
Mistral launches Forge at Nvidia GTC, pitching custom models trained on company data — not just “fine-tune and pray” — for enterprises chasing control and performance.

Announcing Copilot leadership update — Microsoft Blog
Microsoft reshuffles leadership around Copilot and model work, framing it as the next stage: assistants that execute multi-step tasks with “clear user control points.”

Singularity Soup Take: Every org chart update now reads like a battle plan for the assistant era — the prize isn’t a chatbot, it’s the workflow that owns your day.

Safety, Security, and Guardrails (The Boring Stuff That Stops Fires)

AI firm Anthropic seeks weapons expert to stop users from 'misuse' — BBC News
Anthropic is looking for specialist expertise to reduce catastrophic misuse risks — a reminder that “helpful assistant” and “dual-use tool” are roommates now.

AI Security for Apps is now generally available — Cloudflare Blog
Cloudflare says its AI app security layer is now GA, with discovery features aimed at finding and securing “shadow” deployments.

Policy, Rights, and Regulation

AI agents test limits of EU rules — Digital Watch Observatory EU lawmakers and experts are debating liability and accountability for agent-like systems that can act autonomously — because contracts written for humans are feeling very 2019.

Senators tell ByteDance to shut down Seedance 2.0 AI video app 'immediately' — Engadget
US senators urge ByteDance to shut down its AI video generator, citing concerns about copyright and the economic fallout for creators.

Google scraps AI search feature that crowdsourced amateur medical advice — The Guardian
The Guardian reports Google quietly dropped a health-advice search feature, amid scrutiny of AI-generated summaries and misinformation risks in public health contexts.

Compute Arms Race (Chips, Quantum, and Other Expensive Wishes)

UK must learn lessons from AI race and retain its quantum computing talent, says minister — The Guardian
The UK announces a £1bn quantum funding pledge, with ministers warning it can’t keep exporting its best talent (and future companies) to the US.

NVIDIA GTC 2026: Live Updates on What’s Next in AI — NVIDIA Blog
Nvidia’s live blog tracks keynote themes from tokens to platforms, pitching a full-stack “five-layer cake” of the next compute cycle.

From AI agents to orbit: key moments from Nvidia’s annual conference — Euronews A roundup of GTC highlights spanning agents, autonomous vehicles, and space computing — because the only thing better than a data center is a data center in orbit.

3 Questions: On the future of AI and the mathematical and physical sciences — MIT Physics MIT researchers discuss how math and physics both power and are reshaped by modern machine learning — a calmer, smarter counterweight to the hype treadmill.

Infrastructure & Investment

Exclusive: Ireland seeks slice of AI boom — Semafor
Irish companies are chasing opportunities in the US data center buildout, riding the infrastructure wave that keeps “intelligence” fed with electricity and silicon.

Relevant Resources
Google Gemini — What it is, where it shows up, and how it fits into everyday tools.

Microsoft Copilot — A quick guide to where Copilot lives and what it’s meant to do.

Claude (Anthropic) — Context on Anthropic’s assistant and its positioning.

Today's Pulse: 15 stories tracked across 13 sources — TechCrunch, Ars Technica, The Verge, Engadget, BBC News, Microsoft Blog, Cloudflare Blog, The Guardian, Digital Watch Observatory, NVIDIA Blog, Euronews, MIT Physics, Semafor