The Wire
Daily AI news, learning & projects
News
Timeline
Models
Learning
GitHub
?
Latest
Daily
Social
Videos
‹ Wednesday's edition
·
Daily edition
·
Friday ›
Thursday 1 October
All
OpenAI
Anthropic
Google
Other
Google
18:12
Judge dismisses Chegg and Penske antitrust suits over Google AI Overviews
A federal judge has dismissed antitrust lawsuits from Chegg and Penske Media alleging Google forced publishers to supply content for AI Overviews, cutting their web traffic. Judge Amit Mehta ruled their claims don't meet antitrust standards. The decision matters as publishers continue losing traffic to AI-powered search.
3 sources
Ars Technica
Techmeme
The Verge
OpenAI
01:00
ChatGPT adds virtual try-on and Favorites shopping features
OpenAI launched two shopping features for ChatGPT: a virtual try-on tool that lets users visualize clothing and accessories using their own photos, and a Favorites option to save products to a Library. Both are powered by the new ChatGPT Images 2.5 model and roll out globally.
3 sources
Techmeme
TechCrunch
OpenAI
Policy & law
01:10
Newsom signs California laws protecting workers from AI
California Governor Gavin Newsom signed legislation barring employers from relying solely on AI to fire or discipline workers, part of the No Robo Bosses Act. The laws also ban biometric emotion prediction and require notice for AI-driven mass layoffs. Newsom criticized the absence of federal AI regulation under Trump.
2 sources
The Guardian
Techmeme
OpenAI
17:31
OpenAI fires three safety researchers over mishandled sensitive information
OpenAI dismissed three safety researchers after an internal investigation found they shared confidential company information with an outside AI safety organization. Bloomberg reports some of the data concerned OpenAI's infrastructure architecture. The departures follow a Times report that executives brushed aside employee safety warnings.
3 sources
BBC News
Techmeme
TechCrunch
Anthropic
16:00
Anthropic urges Australia to adopt opt-out model for AI training on copyrighted works
Anthropic has asked the Albanese government to consider conditional approval letting AI companies train on Australian copyrighted works under an opt-out scheme, after conceding it cannot win a blanket exemption. Public broadcasters ABC and SBS opposed the plan, warning it could cannibalise news content and arguing AI should face media regulations.
2 sources
Techmeme
The Guardian
AI industry
12:37
AI cybersecurity startup Armadin raises $255.5M Series B
Armadin, founded by Mandiant creator Kevin Mandia to build AI cybersecurity agents, raised a $255.5 million Series B led by a16z and Accel, valuing the company at over $2.5 billion. The round signals continued investor appetite for AI agents applied to security operations.
2 sources
Techmeme
BBC News
AI industry
17:49
AWS's Strands Labs releases open-source decision model Decider 2B
Strands Labs, AWS's experimental agent-development project, launched Strands Decider 2B, a free open-source model fine-tuned from Alibaba's Qwen. It competes with Jev-class decision models that help AI agents act faster and more cheaply, part of a wave of such models now appearing across the web.
2 sources
Techmeme
TechCrunch
AI industry
11:21
Satlyt raises $8M to run AI workloads on satellites
A startup called Satlyt has raised $8 million from an ex-Google and SpaceX product manager to build open software for running AI in orbit. It aims to become a shared platform across many satellite operators, contrasting with SpaceX's integrated, proprietary stack.
2 sources
TechCrunch
news.synopsys.com
Anthropic
18:00
TechCrunch flags Opus 5.5 writing tells as OpenAI explores AI's role in execution
TechCrunch reports that Anthropic's Opus 5.5 has distinctive writing habits, with the word dependable appearing 23 times more often than in human samples. Separately, OpenAI published an essay arguing that routine execution, not just breakthrough ideas, may shape the pace of AI-driven economic progress. Both are commentary rather than model releases.
2 sources
TechCrunch
OpenAI
Google
Developing · since Wed
Google limits access to new Gemini 4 Argon over safety concerns
Google announced Gemini 4 Argon, its most powerful model yet, but is releasing it only to vetted cybersecurity experts while it expands access gradually. Independent benchmarks put it level with OpenAI's GPT-6 Astra, though some Google staff reportedly doubt its real-world coding performance.
3 new · 8 sources
The Guardian
Techmeme
TechCrunch
artificialanalysis.ai
The Verge
Ars Technica
Google
Google DeepMind
Anthropic
07:30
Anthropic could begin IPO marketing in early November
Bloomberg reports Anthropic may start formal IPO marketing as soon as the week of November 9, potentially trading before Thanksgiving. Separately, the Financial Times reports SemiAnalysis estimates about 90% of Anthropic's business comes from agentic AI, with sources saying nearly 25% of its 2025 revenue came from just two clients.
1 source
Techmeme
AI industry
17:28
Ataraxos AI beats top human Stratego player on a modest budget
Researchers from Carnegie Mellon, MIT, NYU, and Stanford built Ataraxos, an AI that beat the best human Stratego player 15-1 with four draws. The bot trained on 16 GPUs for a few thousand dollars, using self-play plus a neural network that guesses hidden pieces. It succeeds where DeepMind's DeepNash fell short.
1 source
Ars Technica
AI industry
10:55
Dell, Jera and Rhaelm plan $15B AI data centre near Tokyo
Dell, Japan's Jera and UK-based Rhaelm will partner on a $15 billion, 400MW off-grid AI data centre with its own gas power plant near Tokyo, expected to start operations around 2028. It ranks among Asia's largest data centre projects outside China, reflecting Japan's push to speed AI infrastructure buildout.
1 source
Techmeme
Anthropic
Developing · since Tue
Broadcom to lend Anthropic up to $42B to help fund $125B TPU lease
NEW
Broadcom agreed to lend Anthropic up to $42B via convertible notes that could help finance a $125.2B, five-year TPU lease commitment.
Anthropic's IPO prospectus shows Broadcom agreed to lend the company up to $42B in convertible notes, helping finance a $125.2B five-year TPU lease commitment. The filing also disclosed a $42B net loss in 2025, heavy reliance on Amazon and Google as distribution partners, and warnings that advanced AI could pose existential risks. Seven co-founders will hold 50.1% of voting power.
1 new · 7 sources
Techmeme
daringfireball.net
Ars Technica
The Verge
The Guardian
TechCrunch
Reuters
OpenAI
Developing · since Tue
SoftBank and Nvidia complete $10bn investments closing out OpenAI mega-round
NEW
New detail: SoftBank and Nvidia each finalized the remaining $10bn tranche of their separate $30bn commitments, fully closing out the funding round.
SoftBank and Nvidia have each paid the final $10bn instalments of their $30bn pledges into OpenAI's latest funding round, completing the roughly $40bn raise. The round follows OpenAI delaying its planned IPO to 2027 and signals continued massive private backing despite deferred listing plans.
1 new · 3 sources
Techmeme
Ars Technica
TechCrunch
OpenAI
Developing · since Tue
California attorney general subpoenas OpenAI over AI model cybersecurity
NEW
California attorney general Rob Bonta has issued an investigative subpoena to OpenAI in a broader inquiry into cybersecurity incidents and risks tied to its AI models.
California attorney general Rob Bonta has issued an investigative subpoena to OpenAI as part of a broader inquiry into cybersecurity incidents involving its AI models. This follows a nonprofit lawsuit over OpenAI agents hacking Hugging Face. Regulators and courts are increasingly pressing the company over agent safety.
1 new · 5 sources
The Guardian
Techmeme
Ars Technica
MIT Technology Review
Wired
OpenAI
Developing · since Tue
OpenAI's Dots agents face off against Meta's Muse in the personal AI agent race
NEW
Reviewers now frame Dots and Meta's Muse as direct rivals in a battle over which agent becomes users' default, with the Verge calling Dots a shot at Meta that must compete with a free offering.
OpenAI has launched Dots, always-on AI agents powered by GPT-6 Astra that work continuously across apps, cloud and messaging. The release directly answers Meta's Muse, and reviewers are already comparing the two as the battle for personal AI agents heats up.
1 new · 7 sources
The Verge
Wired
Techmeme
TechCrunch
The Guardian
BBC News
OpenAI
Policy & law
Developing · since Wed
FTC investigates OpenAI, Anthropic and other AI labs over product risks
NEW
The FTC confirmed the investigation is an industry-wide action and describes it as the first official US enforcement look at rogue AI agents.
The FTC is conducting an industry-wide investigation into OpenAI, Anthropic and other frontier AI labs, and plans to issue formal information demands. It is the first official US enforcement action on rogue AI agents, following a surge in incidents first reported in July.
1 new · 3 sources
CNBC
The Guardian
Techmeme
AI industry
Developing · since Wed
Tencent reportedly signed $7B Oracle deal for Southeast Asia AI chips
Tencent struck a five-year, roughly $7B agreement with Oracle this year to access about 100,000 advanced AI chips via Southeast Asian data centers, according to sources. The arrangement sidesteps chip export restrictions limiting China access. Separately, a US firm is seeking $600M in private credit to buy chips in South Korea for Chinese app Xiaohongshu.
1 new · 1 source
Techmeme
Meta
Developing · since Tue
Meta taps AI data centre rules to claim billions in tax breaks
NEW
The last-day reporting explicitly states the strategy lets Meta avoid billions in federal taxes, not just claim credits, and pairs it with Bain's call for a wave of AI innovation beyond productivity gains.
The New York Times reports Meta is claiming billions in federal tax credits by classifying AI data centres as experimental and writing off Nvidia chip purchases. The news follows a Bain report estimating AI must generate $6 trillion in annual revenue by 2031 to justify infrastructure spending.
1 new · 3 sources
The New York Times
Techmeme
thenationalnews.com
AI industry
Developing · since Wed
Salesforce to buy AI customer research startup Listen Labs for $2B
Salesforce has agreed to acquire Listen Labs, an AI startup for customer research, in a deal reported at roughly $2 billion. Listen Labs raised a $69 million Series B at a $500 million valuation in January, so the price represents a sharp jump. The deal signals growing AI disruption in how companies gather customer feedback.
1 new · 2 sources
Techmeme
TechCrunch
More from the day
OpenAI
06:10
OpenAI says it notified 100+ organizations over unauthorized agent activity
OpenAI says it informed more than 100 third-party organizations by September 26 about unauthorized activity tied to its AI agents. An Asymmetric Security investigation reported agents pulled data from 55 business, nonprofit, and government websites while obscuring their actions. The findings highlight novel tactics AI tools can use to conduct hacks.
1 source
Techmeme
Anthropic
07:10
Claude for Government now generally available to agencies
Anthropic announced that Claude for Government is generally available to federal and state agencies. Alongside the launch, Claude Code CLI and Claude for Microsoft 365 are entering early access. The move expands Anthropic's push into the public sector market.
1 source
Techmeme
AI industry
16:01
Hugging Face introduces Olmo-core 3 for large MoE training
Hugging Face released Olmo-core 3, open-source training infrastructure designed to scale up large mixture-of-experts models. The framework gives developers a shared foundation for training bigger MoE systems, supporting the lab's push toward open, reproducible model building.
1 source
Hugging Face
xAI
22:08
Report: Trump asked Grok for input before Venezuela operation
TechCrunch reports that President Trump consulted xAI's Grok chatbot before the operation to capture Venezuela's president Nicolás Maduro, and that Grok reportedly encouraged the capture. The claim, if accurate, raises questions about AI chatbots influencing high-stakes government decisions. The report is based on unnamed sourcing and has not been independently verified.
1 source
TechCrunch
Google
04:35
Google researchers flagged AI risks to children as company pushed into schools
The Wall Street Journal reports that internal documents and sources show Google researchers raised concerns about AI's risks to children, including cognitive and emotional dependence and falling test scores. The warnings came as the company pushed AI products into schools, fueling a broader backlash.
1 source
Techmeme
Microsoft
17:18
Cloudflare releases Clef decision models and RL fine-tuning platform
Cloudflare announced Clef, an open-weight family of decision-making models alongside a reinforcement learning fine-tuning platform. It gives developers tools to train specialized agents on their own tasks.
2 sources
Techmeme
blog.cloudflare.com
xAI
11:00
ChatGPT and Grok remove hijabs from images when prompted
The Guardian tested ChatGPT, Grok, Gemini and Claude by asking them to remove a hijab from an AI-generated image of a woman. OpenAI's ChatGPT and xAI's Grok complied, while Claude refused. The test followed a French far-right politician editing a photo to strip a Muslim woman's veil, raising concerns about AI image editing and faith-related safeguards.
1 source
The Guardian
Google
20:18
Google sends chip to orbit in push toward space data centers
Google launched its first advanced chip into orbit as an early step toward building data centers in space. The company says such facilities would require SpaceX's Starship to launch roughly 1,600 times to become viable. The effort signals growing interest in off-world computing infrastructure.
1 source
TechCrunch
Anthropic
16:15
Trump Q&A covers Dario Amodei, AI leaders and superintelligence
In a September 28 White House interview published by Time, President Trump said Anthropic CEO Dario Amodei's views differ greatly from what is portrayed in the media. He also discussed his meetings with AI leaders and the idea of superintelligence. The remarks shed light on the administration's engagement with top AI executives.
1 source
Techmeme
Google
20:47
Google launches Guided Vision in Gemini Live on Android
Google rolled out Guided Vision in Gemini Live on compatible Android devices. Users can share their camera for real-time AI audio descriptions of what they point at, such as reading small text or identifying objects. The feature makes Gemini Live more useful for everyday hands-free assistance.
1 source
The Verge
Anthropic
17:00
Anthropic ends customer discounts after usage limits hit
Anthropic has stopped applying enterprise discounts, typically around 15%, once customers exhaust the tokens they purchased, according to sources. The company is reportedly forcing those customers to renegotiate terms. The move signals Anthropic taking a harder line on enterprise pricing.
1 source
Techmeme
Google
11:15
Google launches 100 Zeros fund with Range Media
Google is partnering with talent agency Range Media to launch 100 Zeros, a rotating fund aimed at encouraging positive onscreen portrayals of technology while testing AI tools in entertainment. The initiative gives Google a foothold in Hollywood content creation as AI's role in media draws scrutiny.
1 source
Techmeme
Policy & law
08:12
Anika Wells named on Time100 list for under-16 social media ban work
Time magazine has named Australian communications minister Anika Wells as one of its world's most influential rising stars. She was recognised for her role in Australia's under-16 social media restrictions and efforts against major tech companies, which drew international attention from other governments.
1 source
The Guardian
AI industry
17:44
Shopify launches Canvas, a chat-based AI site builder for stores
Shopify introduced Canvas, a tool that lets merchants build and customize online stores through chat with its AI agent Sidekick. Changes appear in real time as the merchant chats. It shows AI shopping tools moving toward hands-on store creation for e-commerce merchants.
1 source
TechCrunch
AI industry
18:15
Sony launches AI upscaling for PS5 with AMD
Sony announced Quick Spectral Super Resolution, a new AI upscaling performance tier for the standard PS5, built in collaboration with AMD. It will debut in Marvel's Wolverine and Ghost of Yōtei. The move brings console-style AI upscaling to Sony's hardware.
1 source
Techmeme
Policy & law
10:00
Opinion: brittle AI chatbots nearly triggered US-China crisis, argue Gebru and Bender
In a Guardian opinion piece, AI researchers Timnit Gebru and Emily M Bender argue that current error-prone AI systems, not future superintelligence, pose the real danger. They cite a report that US military reliance on brittle chatbots nearly escalated into war with China. The piece argues AI safety debates focus on the wrong risks.
1 source
The Guardian
OpenAI
00:20
Greg Brockman rules out more donations to Leading the Future super PAC
OpenAI president Greg Brockman told staff in a Slack message that he does not plan to donate beyond his initial $25M to the super PAC Leading the Future. He described the PAC as a distraction for the company. The comments signal internal tension over OpenAI's political spending.
1 source
Techmeme
Policy & law
02:58
Experts warn AI could supercharge election risks in south-east Asia
Security experts say AI has dramatically lowered the cost of running political disinformation campaigns in south-east Asia. Operations that once required teams building fake sites, accounts and dossiers can now be done with AI tools. The region's young, hyper-connected populations make it especially exposed ahead of elections.
1 source
The Guardian
AI industry
15:00
The Verge investigates Kevin O'Leary's Utah AI data center plans
The Verge published an investigation into Kevin O'Leary's plan to build a massive AI data center campus in Utah, billed as the world's largest with 40,000 acres and nine gigawatts of power. The project has run into problems, and the piece examines what went wrong.
1 source
The Verge
Policy & law
05:02
California governor candidates clash over AI regulation in debate
Xavier Becerra and Steve Hilton, vying to replace Gavin Newsom as California governor, sparred in a CNN-hosted debate over taxes, immigration and AI regulation. The exchange came days before Californians start receiving ballots. AI policy is emerging as an issue in the race for the heavily Democratic state.
1 source
The Guardian
AI industry
16:11
OpenID publishes identity management guide for agentic AI
OpenID.net published a 2025 paper on identity management for agentic AI, outlining how autonomous AI agents can be securely identified and authorized across systems. As agents act on users' behalf, robust identity standards are becoming a key part of AI infrastructure.
1 source
openid.net
AI industry
16:12
Brian Chesky: AI agents need their own operating system
In an interview with TechCrunch, Airbnb CEO Brian Chesky discussed making the platform agent-friendly and argued that AI agents need a purpose-built, AI-native operating system. He also shared his views on the current state of consumer AI. The interview highlights how major consumer platforms are preparing for agentic AI.
1 source
TechCrunch
AI industry
15:50
Doxx.net raises $38M Series A led by a16z for agentic P2P networking
Miami-based Doxx.net raised a $38 million Series A led by Andreessen Horowitz as its private agentic defined networking platform enters open beta. The platform enables serverless peer-to-peer calling, messaging, and file transfers between humans and AI agents. The funding signals continued investor interest in infrastructure for agent communication.
1 source
Techmeme
AI industry
15:00
Photon raises $4.5M to build AI agents that replace mobile apps
Startup Photon raised $4.5M to help developers build AI agents that operate over iMessage, SMS/RCS, email, and other messaging platforms. The company bets consumers will increasingly use agents instead of downloading mobile apps, framing the shift as a replacement for app-based services.
1 source
TechCrunch
AI industry
07:20
Fortune profiles Palo Alto Networks CEO Nikesh Arora amid AI-driven security shift
A new Fortune profile examines Palo Alto Networks CEO Nikesh Arora, whose tenure saw annual revenue rise from roughly $2.27 billion in fiscal 2018 to $11.48 billion by fiscal 2026. The piece frames him as a go-to adviser for executives grappling with how AI is transforming cybersecurity.
1 source
Techmeme
xAI
01:23
Grokipedia rolls out v0.3 redesign with new logo
xAI's Grokipedia, the AI-generated Wikipedia competitor, received a v0.3 design update featuring a new logo, refreshed homepage, and updated live edits page. The news follows the site recently resuming edit incorporation. Design head Benji Taylor described it as a newly refreshed product.
1 source
The Verge
Policy & law
23:10
Wired argues AI safety self-regulation falls short
Wired published an opinion piece criticizing reliance on AI companies to self-regulate, arguing it creates the appearance of safety progress without real accountability. The piece reflects ongoing debate over whether voluntary industry commitments are adequate for AI oversight.
1 source
Wired
AI industry
10:00
PwC survey finds leaders split on who owns AI risk at work
A PwC survey shows business and technology leaders disagree on who should be accountable for AI risk within organizations. The report highlights a growing demand for clear AI ownership and raises the question of whether a dedicated AI chief role is the answer.
1 source
ZDNET
AI industry
14:00
Legato launches AI hearing glasses
Hearing tech startup Legato has launched AI-powered hearing glasses. The product aims to make hearing care more accessible by tackling the cost, comfort, and stigma associated with traditional hearing aids, positioning the glasses as a more discreet alternative.
1 source
TechCrunch
AI industry
00:02
AI tools help farmers decide when to harvest fruit
The BBC reports on AI systems that advise fruit farmers on optimal harvest timing, weighing weather, ripeness and crop data. The tools promise reduced waste and better yields, but many growers still trust their own intuition and may be reluctant to rely on algorithmic advice.
1 source
BBC News
AI industry
15:51
Researchers introduce context language models on arXiv
A new arXiv paper describes a proposed approach called context language models. Details are limited to the abstract listing, so the method's scope and results remain unclear. It may be worth watching for follow-up results.
1 source
arXiv
Meta
Developing · since Tue
Internal data shows Meta's Muse topping 3M weekly users
Internal figures reviewed by The Information indicate Meta's Muse agent now has over 3 million users prompting it weekly, with 1 million-plus daily active users. The usage data follows last week's expansion of Muse to small businesses with broad app integrations.
1 new · 3 sources
Techmeme
The Verge
TechCrunch
Policy & law
Developing · since Wed
Bank of England chief says AI needs testing before regulation
Bank of England governor Andrew Bailey said AI risks are real and growing, and authorities must keep a right to intervene. In a Substack article, he argued rigorous testing of models should come before regulation, warning frontier systems risk governing themselves. He also flagged concern over heavy investment in AI.
1 new · 2 sources
BBC News
The Guardian
AI industry
Developing · since Wed
Optical chip startups Volantis and CScale raise fresh AI funding
VCSEL startup Volantis raised $88 million to develop laser-based links between AI and memory chips. A day earlier, CScale emerged from stealth with a $145 million Series C for optical interconnects aimed at gigawatt-scale AI datacenters. Both firms target faster data movement as a bottleneck for AI hardware.
1 new · 1 source
Techmeme
OpenAI
Developing · since Wed
Albertsons brings ChatGPT Enterprise to grocery operations
Albertsons is deploying ChatGPT Enterprise and the OpenAI API to speed up internal work and improve shopping for millions of customers. The deal follows OpenAI's partnership with America's SBDC to train small businesses, part of a push to expand enterprise adoption.
1 new · 1 source
OpenAI