AI Reading for Monday January 12

Google unveiled the Universal Commerce Protocol, an open standard developed with partners including Shopify, Etsy, Wayfair, Target and Walmart, interoperating with agent protocols and enabling direct AI-mode checkout in Search and Gemini for U.S. retailers (Google Pay now, PayPal soon). - TechCrunch

Google launched the Universal Commerce Protocol (UCP), an open standard co-developed with Shopify, Etsy and Walmart, adding in-Search checkout, a Business Agent for brand-voiced help with early adopters Lowe's, Michaels, Poshmark and Reebok, and Direct Offers. - Engadget

Google introduced new AI agents for retailers to power AI-based shopping. - The Wall Street Journal

AdSense and Sensibility: Google is introducing personalized shopping ads and product suggestions into its AI tools, integrating shopping ad placements into AI-driven interfaces to monetise popular AI features and boost advertiser targeting and e-commerce conversions. - Financial Times

European Commission President Ursula von der Leyen condemned Elon Musk's platform X for sexually explicit deepfakes produced by its AI chatbot Grok, saying the tool has sparked investigations and warning the EU may impose new child-protection and consent rules. - POLITICO

Ofcom opened a formal investigation into Elon Musk's X to determine whether its Grok AI chatbot was used to create and share non-consensual intimate and sexualised images, including alleged child sexual abuse material, potentially breaching UK law. - Yahoo

The humiliation is the point: Women speak out over Grok’s nonconsensual sexualized images of women. - Newsweek

Epic Games CEO, known for fighting to open app stores, says Grok backlash is crony capitalism pushing down a political opponent - The Gamer

I think this is a questionable slippery-slope argument, I don’t think there is any choice but to have guardrails against awful stuff, chemical/bioterror recipes and whatnot, democracy means deciding what is protected speech and what is a nexus to harm, while protecting fundamental rights.

there are sketchy prompts that an innocent person could put in, or a 12-year-old boy without understanding the gravity. - ChatGPT

If you have a button that says 'press here for CSAM of your niece', you don't really get to say, "it's just a tool, you could do it with pastels or photoshop". Responsible online platforms have a responsibility to implement guardrails against harmful stuff. If the market doesn’t speak against 4chan stuff, it’s not crazy for authorities to specify standards.

UK, Canada and Australia are said to be in talks to yeet X after users exploited Grok to generate explicit deepfake images of women and children. - GB News

No, Canada is not considering banning X in the wake of the company's Grok decisions that led to widespread AI-generation of CSAM & NCII etc. (Elon Musk's xAI owns and operates X and developed and operates Grok.) - CityNews Ottawa

All-seeing AI watches over you while you sleep: SleepFM, published in Nature Medicine (Jan 2026) and co-led by Stanford, was trained on ~600,000 hours of polysomnography from ~65,000 sleepers and can use one night of brain, heart and body signals to predict risk for ~130 diseases. - dw.com

Anthropic launched Claude for Healthcare with HIPAA-ready infrastructure, expanded Claude for Life Sciences, upgraded model Opus 4.5, and added connectors to CMS, Medidata and ClinicalTrials.gov while stating user health data will not be used to train models. - Anthropic

Anthropic launched Claude for Healthcare, a HIPAA-ready product built on Claude Opus 4.5 that adds connectors to the CMS Coverage Database, ICD-10, NPI Registry, PubMed, clinical trial and life-sciences data, plus tools for prior authorization and FHIR app development. - Business Insider

People are big mad at Anthropic for blocking monthly plans from open-source coding assistants, feels like a bait-and-switch. - Archaeologist

But Anthropic is making bank from the API for enterprise coding clients using Cursor, who can't use e.g. Max plans, they can ill afford to give OSS coding assistants a huge discount.

If you think the Claude Code CLI is the same as e.g. Cursor, you could argue it's anti-competitive to give their own CLI an edge. OTOH the CLI agent is a different animal, and that's how they get data to train it. Expecting them to let competitors train at the same discount seems unrealistic.

Anthropic unbans some users whose usage of Claude Code auth via alternative agent harnesses triggered abuse flags - X (formerly Twitter)

B-Ilya-naire: ext messages from OpenAI COO Brad Lightcap, released in Elon Musk's lawsuit, show Ilya Sutskever held about $4 billion vested equity in November 2023, which would equal roughly $117 billion at an $850 billion valuation. - ctech

quite a last laugh, regardless of what happens to his startup

Harmattan AI raised $200 million at $1.4 billion valuation in a Series B led by Dassault Aviation, to scale production of 10K surveillance and electronic warfare drones per month. - TechCrunch

See something, say something: MTA issued a request for information to explore AI video analytics across more than 15,000 cameras and 6,000 subway cars to detect weapons and monitor items, while critics cite a 2024 pilot's 12 knives and 100+ false positives. - THE CITY - NYC News

but can it detect an emotional disturbed person’s intrusive thoughts about pushing you off the platform

Reuters Institute's survey of 280 media leaders across 51 countries found publishers expect search referrals to fall about 43% over three years, prompting shifts toward subscriptions, direct audience relationships and short-form creator-style content and partnerships. - the Guardian

AI industry insiders launch "Poison Fountain" project to let websites return bad data when AI scrapers visit. - The Register

big tech pushes poison, well two can play that game

Silicon Valley's vision of AI-driven economic abundance risks concentrating benefits with the corporations and investors that own the technology, rather than distributing gains broadly. - The New Yorker

Broader integrated software platforms for business will win out over vertical start-ups and solutions targeting one layer in the stack. - Financial Times

Job Seekers Find a New Source of Income: Training AI to Do Their Old Roles - The Wall Street Journal

Aurum Capital Management launched WealthOS, an AI-driven quantitative wealth advisory platform whose development began in 2019, with a positive internal review of v4.0 in 2024 and v4.5 in 2025 adding generative AI, reinforcement learning, multi-account support and improved explainability. - Private Banker International

Lego launched a Computer Science & AI curriculum for K–8 classrooms debuting in April, selling $339.95 kits for groups of four across three grade bands that run offline and, Lego says, keep child data on-device. - Fast Company

AI vision can be subject to the same optical illusions as human vision. - BBC

Trump Administration Wants to Achieve ‘Pax Silica’ Through AI. Here’s What That Means - Gizmodo

Means they are fans of the Roman Empire, imperial overreach, good luck blocking everyone else from getting AI

Mask slips, Musk demands UK regime change: "There's gotta be a change of government in Britain and we don't have another 4 years or whenever the next election is. There's gotta be a dissolution of parliament." - X (formerly Twitter)

I seem to remember him saying Twitter must be neutral tho? - The Guardian

Musk and Grok are your government now. If a guy is a anti-democratic racist POS fomenting anti-immigrant riots and state campaigns of terror, it’s not crazy to think about how to put him on a tighter leash.

even some of the card carrying bootlicker caucus seems to be getting it - Noahpinion Blog

Follow the latest AI headlines via SkynetAndChill.com on Bluesky