AI News from Wired AI
Latest coverage from Wired AI, summarized and scored for signal.
- Here's Why You Shouldn't Use AI to Vote — On US Election Day, WIRED reports that mainstream chatbots still give wrong or partisan-sourced voting answers, with Forum AI finding nearly 40 percent of election responses contained at least one factual error.
- ‘Our Risk, Their Reward’: Zohran Mamdani on AI and Data Centers — In a WIRED interview ahead of the midterms, New York City mayor Zohran Mamdani discussed data centers, AI in schools, and the Democratic party's future, framing data center growth as public risk for private reward.
- When You Should—and Shouldn’t—Use AI to Search the Web — Wired explains when AI chatbot search is reliable and when it isn't: fine for widely known facts and complex queries, risky for obscure questions where it may pick the wrong source and still answer confidently.
- Hundreds of Mundane AI Apps Are Hiding Sexually Abusive Deepfake Features — WIRED found a network of about 400 apps and domains, linked to Chinese-language developers, that pose as ordinary AI tools but secretly load a sexually abusive deepfake generator, distributed via Apple and Google stores and promoted with explicit ads on Facebook and Instagram.
- Chatbots Are Leading Hikers Astray, With Dangerous Consequences — Rescuers on Mount Shasta and in other parks report hikers getting into trouble after relying on chatbots like Gemini and Claude for trip planning, a sign of over-trust in AI advice in high-risk settings.
- Spooky New Infostealer Malware Uses AI to ‘Autodox’ Victims — Cisco Talos found Volta-Autodoxer, an infostealer that sends stolen victim data to commercial LLMs to build citation-backed dossiers and rank the most valuable targets.
- Meta’s Muse Could Upend All Kinds of Businesses. It Hasn’t Yet — Despite Wall Street fears about Meta's Muse shopping agent, executives at about 50 companies told WIRED and investors that agent-driven orders are still tiny.
- An AI Agent Injected Drugs Into a Live Mouse’s Brain — Exin Therapeutics let an AI agent pick a gene therapy and control a robotic rig to inject it into a live epileptic mouse's brain, an early test of agents operating lab equipment directly.
- How a Contractor’s Claude Agent Got Into the Pentagon — WIRED's Uncanny Valley podcast covers its scoop that a defense contractor connected Claude Code to a secure Department of Defense system, plus takeaways from Anthropic's S-1 IPO filing.
- New York Seizes a Dozen Celebrity Deepfake Websites — The Manhattan DA's Office seized 12 nonconsensual deepfake pornography websites targeting ~1,200 victims — the largest such enforcement action to date — using the US Take It Down Act, which now gives law enforcement authority to seize domains hosting synthetic sexual imagery.
- AI Leaders Are Calling for a Slowdown. Trump’s Team Says It’s on Them — Anthropic CEO Dario Amodei publicly called for the US government to help slow AI development and coordinate internationally; Sam Altman and Elon Musk backed the call, but Trump's AI advisor David Sacks rejected it, telling labs to act unilaterally rather than seek antitrust cover for a cartel.
- Sexually Explicit Deepfake Sites Target 100-Plus Politicians in Europe — A study of 160 deepfake abuse websites found 138 female MPs from 22 EU countries targeted with nonconsensual explicit content — women are 33 times more likely to appear than male MPs, confirming deepfake porn as a systematic tool to push women out of public life.
- ‘I Like My Big Rat Wife’: Meet the People Using Chatbots to Write Custom Fiction — A University of Washington study analyzing 500,000+ ChatGPT prompts from the WildChat dataset found that more than 1/3 of chatbot interactions involve fiction generation — fan fiction, erotica, role-play — driven largely by power users who treat AI as a collaborative storytelling engine, not an author substitute.
- AI Agents Are Thirsty for Power — The shift from chatbot queries to agentic AI — where systems self-prompt hundreds of times per task — is the core driver behind Big Tech's data center boom; OpenAI's 10,000-agent math swarm generated 2.7 million messages, likely costing tens of millions of dollars in compute for a single run.
- From Hacks to Bioweapons, Claude Misuse Is Now Everywhere — Wired's final security roundup covers Claude being misused for cyberattacks and bioweapon research, Meta failing to remove roughly 350 AI-generated child abuse ads, and Clearview AI testing a new tool that maps a target's associates and social media presence.
- Meta Sued Over Training Data for Its AI and Face-Recognition Systems — A federal class action filed in Chicago alleges Meta illegally harvested Facebook and Instagram photos—without consent—to build NameTag face recognition for its Ray-Ban smart glasses and to train Emu and Muse Image AI models, violating Illinois and California biometric privacy laws.
- One of AI’s Fiercest Critics Says All the Doom Talk Is ‘Meant to Distract Us’ — Timnit Gebru argues that AI extinction fears promoted by labs like Anthropic are a deliberate distraction from concrete present harms — autonomous weapons, surveillance, and labor exploitation — and that doom narratives serve a regulatory capture agenda benefiting the companies stoking the panic.
- Why So Many AI Researchers Think the Machines Could Kill Everyone — A senior Anthropic safety leader publicly stated a greater than 10% personal probability that AI kills all humans within a decade, as researcher resignations from DeepMind and Anthropic over recursive self-improvement risks reach what Wired calls a fever pitch.
- OpenAI Wants to Know if an AI Industry Slowdown Would Even Be Legal — OpenAI is asking Congress whether coordinating an industry-wide slowdown in frontier AI development would violate US antitrust law, after the company's chief scientist publicly called for voluntary slowdowns as a safety measure. A bipartisan bill introduced in July would grant labs legal cover but has not cleared committee.
- Everything New You Can Do With Siri AI — Apple's iOS 27 introduces Siri AI — a fully rebuilt assistant with on-device personal context indexing, screen awareness, and native text generation that drops the ChatGPT integration dependency. Available only on iPhone 15 Pro or newer, it marks Apple's most direct push yet against dedicated AI assistants like ChatGPT and Google Gemini.
- Clearview AI Is Testing an AI Tool That Would Let Cops Unearth Your Life Online — Wired discovered Clearview AI has quietly built InquiryIQ, a prototype AI analyst tool that takes a facial recognition match and auto-generates an online profile — employers, aliases, associates — using xAI's Grok. The tool accepts age, gender, and race inputs to shape searches and was found embedded in Clearview's own login page JavaScript.
- The AI Researcher Who Just Quit Anthropic Says It’s ‘Crunch Time for Humanity’ — Anthropic pretraining researcher Jacob Coxon resigned publicly with a 100M-view X post warning that colleagues describe the current moment as "crunch time for humanity" — and Anthropic's own alignment lead separately posted a greater than 10% probability that AI kills all humans within a decade, with current and former lab researchers expressing agreement.
- San Francisco Orders Meta to Stop ‘Allowing’ AI Child Abuse Ads — San Francisco's city attorney sent Meta a cease-and-desist after the company ran 350+ paid ads featuring AI-generated child sexual abuse imagery on Facebook and Instagram — including depictions of confirmed real children — with 250+ ads continuing to run even after Wired reported the first batch in August.
- I Let an AI Agent Hack All My Gadgets—and I’d Do It Again — A Wired reporter let an abliterated GLM-5.3 agent (Abliteration AI) loose on their home network; it found vulnerabilities in household devices, hacked into a PC, exposed bugs in vibe-coded projects, and generated a prioritized hardening plan—suggesting defensive AI value may exceed offensive risk for most users.
- A Stealth Startup Thinks It Just Hacked the Memory Shortage — Kepler Computing exits stealth with $468M raised — including $245M from the US Department of Commerce — claiming its 3D-stacking approach builds high-bandwidth memory without expensive EUV lithography, directly targeting the HBM shortage bottlenecking AI data centers.