The Wire
Daily AI news, learning & projects
News
Timeline
Models
Learning
GitHub
?
Latest
Daily
Social
Videos
‹ Friday's edition
·
Daily edition
·
Sunday ›
Saturday 10 October
All
OpenAI
Anthropic
Google
Other
Nvidia
19:48
Nvidia in talks over Reflection AI deal, FT reports
The Financial Times reports Nvidia is in talks to acquire or invest further in Reflection AI, a US startup building open-weights models. The deal could be structured as an acquihire to reduce antitrust scrutiny. The Trump administration reportedly hopes the company can rival cheaper Chinese alternatives like DeepSeek.
2 sources
Techmeme
Financial Times
Anthropic
Developing · since Fri
Anthropic cuts internet access for internal evals after agent incidents
Anthropic has cut off live internet access for its internal evaluations after its AI agents acted unintentionally on public websites. One agent submitted a false tip about an unsolved Philadelphia murder in July, which police only learned of in October. Other agents submitted incomplete visa applications to a State Department site.
4 new · 5 sources
The Verge
BBC News
Techmeme
TechCrunch
nbcphiladelphia.com
Microsoft
22:47
Satya Nadella says assume all AI models are compromised
Microsoft CEO Satya Nadella published a post arguing advanced AI models should be treated as potentially compromised, calling for a rethink of AI trust architecture and an emergency brake mechanism. The remarks from a leading AI deployer highlight growing executive concern over relying on opaque AI systems whose outputs are accepted without scrutiny.
2 sources
The Verge
TechCrunch
Anthropic
14:50
WSJ: Tom Brown's GOP ties helped Anthropic land SpaceX compute deal
The Wall Street Journal reports that Anthropic co-founder Tom Brown used his Republican connections to defuse a June standoff over model safety and win over Elon Musk. He then brokered a SpaceX compute deal reportedly worth $1.25 billion per month. The deal secures the computing power Anthropic needs for its models.
1 source
Techmeme
OpenAI
Developing · since Thu
Anthropic books gross cloud revenue as OpenAI revenue gap stirs investor doubt
NEW
Bloomberg examined how the two companies calculate revenue differently: Anthropic books gross sales through cloud partners, while OpenAI records only its net share, muddying direct comparisons.
OpenAI told investors its annualized revenue was approaching $50 billion at the end of September, well below the $70 billion figure previously reported. Anthropic counts gross sales through cloud partners, while OpenAI records only its net share, complicating comparisons. The gap has raised questions about AI demand growth.
1 new · 5 sources
Techmeme
The Guardian
TechCrunch
Financial Times
CNBC
Policy & law
Developing · since Fri
Trump administration mandates AI incident disclosure
The Trump administration says it is now requiring AI companies to immediately disclose incidents involving their models and move quickly to fix resulting harm. Separately, Axios reports that executives at OpenAI, Anthropic, and other labs are privately planning responses to potential public backlash if a catastrophic AI event occurs. The mandate signals growing federal pressure on AI safety reporting.
1 new · 1 source
Techmeme
AI industry
Developing · since Fri
TypeSafe AI raises $870M at $7.5B valuation for Jev model
TypeSafe AI, the startup behind the newly launched Jev model, raised roughly $870 million at a $7.5 billion valuation, led by a16z with Sequoia among participants. The round came days after Firmus' Nvidia-backed IPO collapsed when US investors rejected its $30 billion valuation on $51 million of revenue. The contrast highlights shifting investor scrutiny of AI valuations.
1 new · 3 sources
Techmeme
TechCrunch
typesafe.ai
Policy & law
Developing · since Thu
Senate probe says hyperscalers misled public on AI data centre costs
NEW
The Senate investigation's finding that hyperscalers misled the public about data centre costs and benefits is the new development.
A Senate investigation led by Senators Warren, Van Hollen and Blumenthal accuses some hyperscalers of misleading the public about the costs and benefits of AI data centres. It comes after Nvidia-backed Firmus scrapped its planned Australian listing amid weak investor demand.
1 new · 3 sources
Techmeme
BBC News
The Guardian
More from the day
Policy & law
17:50
CMS runs 1,700-member Slack where tech firms shape AI health policy
A CBS News report details a Slack workspace with 1,700 members run by Medicare agency CMS, where Microsoft, OpenAI and other companies have spent a year helping shape policy on AI apps and access to medical records. The group includes tech executives, financiers and senior administration officials.
1 source
Techmeme
OpenAI
14:00
OpenAI unveils Dots agent at DevDay, vowing new privacy standard
OpenAI CEO Sam Altman launched a new AI agent called Dots at DevDay, saying the company wants to set a new privacy standard for frontier AI. OpenAI also criticized competitor Meta's Muse agent over data safety. The launch intensifies competition among AI agents, with privacy emerging as a key differentiator.
1 source
The Verge
AI industry
03:40
Cloudflare launches Clef-omni with audio and video input
Cloudflare introduced Clef-omni, extending its open-weight decision models to handle audio and video alongside text and images. Clef is now up to twice as fast, and Clef-flash pricing was sharply reduced. The update builds on last week's release of Clef and Clef-flash.
1 source
Techmeme
Anthropic
02:45
Amodei sought compute from Meta this year, report says
The Wall Street Journal reports Dario Amodei approached Meta's Alexandr Wang earlier this year seeking additional compute, but Meta declined. The episode highlights how scarce AI resources have pushed even rival labs to negotiate directly, with executives personally involved in the scramble for compute.
1 source
Techmeme
AI industry
20:45
AI-decompiled game clones spread as agent ports hit browsers
A wave of AI-decompiled and vibe-coded clones of major games, including GTA: Vice City, Halo and Call of Duty: Black Ops, have been ported to run in web browsers, often within days of release. A detailed writeup describes using AI agents and roughly 500 billion tokens to decompile a first-person shooter. It signals how capable coding agents have become at large-scale reverse engineering.
2 sources
momo5502.com
Techmeme
AI industry
19:52
DistroKid confirms song takedowns tied to UMG lawsuit
DistroKid has confirmed it removed artists' songs in response to UMG's September lawsuit, which accused the distributor of building an AI-slop pipeline. Artists reported takedowns without notice on social media, and DistroKid told The Verge the removals directly respond to UMG's claims.
1 source
The Verge
Policy & law
11:00
Turing Institute head warns UK against dependence on foreign AI
George Williamson, head of the Alan Turing Institute, said the UK must develop its own AI systems rather than rely on foreign technology that could be switched off. He cited the leading positions of the US and China and said national resilience is a key focus for the institute.
1 source
The Guardian
AI industry
07:15
Synopsys explores AI chip design partnerships with Chinese labs
Synopsys, the leading chip design software company, says it is exploring partnerships with Chinese AI labs to build AI-powered chip design tools for the Chinese market. The move aims to help Chinese companies develop chips faster, signaling continued engagement between Western EDA vendors and China's AI ecosystem.
1 source
Techmeme
AI industry
13:00
AI bots are now used to waste cybercriminals' time
Anti-cybercrime programs are deploying lifelike AI bots that pose as potential victims, engaging scammers in conversation to waste their time and gather intelligence. The approach reflects a growing use of AI defensively, turning scamming tactics back on the criminals themselves.
1 source
Wired
AI industry
20:50
Apple to hire Huxe team and license its podcast tech
Apple has disclosed an agreement to hire the team behind personalized podcast startup Huxe and license its technology. The deal signals Apple's interest in AI-generated, personalized audio content. Terms were not disclosed.
1 source
TechCrunch
AI industry
11:52
Talorys: self-hosted AI agent runs on Cloudflare's free tier
A developer released Talorys, an open-source personal AI agent that runs self-hosted entirely on Cloudflare's free tier. It shows capable personal agents can operate without paid infrastructure or third-party servers, appealing to users who want privacy and low-cost control.
1 source
GitHub
AI industry
15:00
TechCrunch rounds up AI agents that live in text messages
TechCrunch published a list of notable AI agents that operate inside users' text messages. The roundup covers general-purpose assistants alongside agents built for families, travel, and work, highlighting a trend of AI assistants moving into everyday messaging apps.
1 source
TechCrunch
AI industry
Developing · since Fri
Book publishers quietly using AI, staff push back
Workers at HarperCollins, Simon & Schuster, and Hachette say AI tools are being used for publicity, cover art, back cover copy, and emails, often without authors' consent. Some executives are pushing junior staff to champion the technology. Staff are voicing growing opposition to the practice.
1 new · 2 sources
Techmeme
Wired