Section
Reporting on Artificial Intelligence.
Reuters, citing the Financial Times, describes Anthropic commitments covering 1.7 gigawatts of heavy-duty gas generators from OnePWR Solutions; independent confirmation remains limited.
Two tech outlets relayed Bloomberg reporting that Apple aims to make Siri track and continue goals more autonomously, with one journalist forecasting a possible shift as early as 2027.
Anthropic’s November 6 engineering post by Lance Martin defines agent memory as retained readable context and describes background dreaming that reflects on logs to correct notes and build linked structures, based on internal evaluations.
Meta describes Spaces as persistent, user-described interactive workspaces inside Muse, distinct from one-off answers, with sharing still exploratory.
The Qwen3.8-27B fine-tune changes responses on Nous’s own tests; the results do not establish neutrality or universal factual accuracy.
Meta reports internal gains on memory and focus for unnamed models reaching people who use Muse and Meta AI, without disclosing the training method or rollout details.
Bloomberg reported that OpenAI closed an initial roughly $11 billion of a funding round valuing the company at $1.25 trillion, with a larger total still prospective and no company confirmation.
The Financial Times reports plans to list before Thanksgiving at roughly $1.5 trillion, while earlier Reuters reporting describes investor doubts about timing and valuation.
The reported recruitment remains an intelligence allegation, while a documented House request seeks a November 18 briefing. Nvidia disputes sharing sensitive information.
The holding is a quarter-end snapshot, not evidence of a fresh cash investment. Separate Qualcomm filings document shares issued to Alphabet-affiliated GV funds in the Modular acquisition.
The collaboration expands Hacktron’s frontier-model access via Daybreak Red while targeting how OpenAI systems connect to external services as AI agents proliferate.
Anthropic’s Frontier Red Team reports Claude made profits in a simulated EVE Online market and coordinated in live EVE Frontier, while instructed saboteurs were rarely caught.
Physical Superintelligence says an internal system using a non-public Anthropic model produced a Lean-accepted proof of the 2008 quantum-query conjecture; a public repo is checkable, but key caveats remain.
The team heads to Meta Superintelligence Labs, the price is undisclosed, and Conductor Cloud shuts down Nov. 20.
Pangram’s new detectors map AI content inside images and check video frame by frame; all accuracy figures are the company’s own.
OpenAI’s consumer product lead says ChatGPT added 200 million weekly users in about 60 days. The search comparison is OpenAI’s own, unpublished estimate.
The five-year deal would have given Tencent access to about 100,000 advanced AI chips in Southeast Asia, according to the Financial Times.
The startup’s review tools already power parts of the Codex app. Axios frames the deal as a challenge to Microsoft’s GitHub.
Tencent says weights and system cards will follow in the coming days. Its benchmark results come from its own internal evaluations.
Wu, who also oversaw Cowork, gave no reason for her exit, which comes weeks before Anthropic’s expected IPO.
P-H3-Image renders in 8 steps and fits a 21 GB VRAM budget at 4-bit, but its license excludes the U.S., EU, UK and South Korea.
The anonymous model takes text, images, video, files and audio and supports tool calling. Speculation that Meta built it is unconfirmed.
Uber’s president will lead OpenAI’s global communications and report to Greg Brockman, a day after the company hired former Army Secretary Dan Driscoll.
Developers can start connecting apps to Siri AI in the 27.3 betas; customers get it “in a few months.”
U.S. users can talk through spending, subscriptions and credit questions aloud, but ChatGPT still cannot move money.
Investing.com, citing The Information, reports that a 15-person Anthropic hardware unit is using Claude to speed custom silicon work.
Bessemer Venture Partners led the round for the generative media platform, which Bloomberg reports doubled its valuation since April.
Some London-based staff want lifetime non-disparagement language dropped from their contracts, The Washington Post reports.
OpenAI describes a device designed with GPT-6 Astra, while a separate Google Cloud benchmark shows how far agents remain from reliable end-to-end engineering.
A new Messages API beta lets agents send only what's new each turn, cutting time to first token without changing token usage.
Anthropic's open-source research prototype uses cryptographic proofs to show which model produced a response.
TestingCatalog reports an unannounced model identifier in Codex usage analytics; OpenAI has not confirmed any GPT-6.2 Astra model.
The indie game platform frames the purge as spam enforcement, not a ban on AI-made games, and is adding Generative AI disclosure filters to search.
The Financial Times reports a verbal requirement for new contracts that Nvidia has neither confirmed nor denied.
OpenAI says the Codex-built system now runs its whole monorepo, cutting median push latency by nearly 5x and serving nearly a petabyte of Git data a day.
In a September 16 post, Sentient described how its coach model turned a grader flaw into a reusable skill.
Critical bugs in core products now pay up to $30,000, exceptional findings up to $172,000 with bonuses, and security firm Cathedral joins OpenAI's next sprint.
The deal would join Cresta's voice agents and real-time agent coaching with Service Cloud; The Information puts the price at $2.7 billion.
Developers ask plain-language questions about raw audio and get back probabilities, priced at $0.47 per million input tokens.
Researchers can query single cell data in plain language while a Claude-powered agent plans, runs and documents the analysis.
Big buyers cite the tech selloff and AI-spending doubts, The Wall Street Journal reports. Anthropic has not set a valuation, share count or date.
Internal guidance updated in October restricts Claude Code and Microsoft's GitHub tools after a productivity and risk review, Bloomberg News reports.
An edit to 0.016% of Qwen3.5-0.8B's weights lifted its MATH-500 score from 7.9% to 48.1%, according to a blog post from MIT CSAIL researchers. The results are not peer reviewed.
U.S. commercial revenue grew 148% to $984 million and passed U.S. government revenue for the first time, and Palantir raised its full-year guidance.
OpenAI's chief named energy as his biggest challenge, insisted humans must stay in control, and repeated a Navier-Stokes claim the Clay Institute has not recognized.
The Sequoia-backed startup says user corrections and edits can feed weekly model updates, with every change gated by the customer's own evals.
Viewers spotted the AI-made short by its clashing art styles, and one asked whether a boy in it was Ben from Cartoon Saloon's Song of the Sea.
The amended S-1 reportedly tells investors that internal estimates point to future models soon meeting a safety-policy threshold for AI-accelerated progress. No current model has been classified as past it.
The oil giant puts about SAR 9.9 billion in cash plus AI and computing assets into HUMAIN. PIF keeps 51% as the company targets 3 GW of data center capacity by 2030.
Apps sold as pet portraits and photo editors linked to a network of 400+ domains offering "nudify" tools. Apple, Google and Meta removed many after WIRED's inquiries.
An internal memo seen by The Register says the need for some roles has "greatly diminished," and affected staff will be told by November 4.
An Anthropic employee says two sentences were added to Claude Code's Opus 5.5 prompt to steer it away from recreating existing games; the cause of separate reverse-engineering refusals is unconfirmed.
Claude Mythos 5, Opus 4.7, Opus 4.6 and Sonnet 4.6 start carrying the mark on November 3; xAI turns on watermarking for seven Grok API models on November 7.
Users describe a tool in plain language, and the agent writes, installs and checks the plugin inside DeepSeek Harness, which remains in public preview.
OpenAI says GPT-6 Astra averages 80.6% on its everyday-tasks benchmark, up from 49.3% for GPT-5.2, but has not released the tasks or the grading method.
The a16z-led Series B funds model training plus more Sonny and Cassie robots, with a goal of more than 1,000 robots under contract by the end of 2027.
Sam Altman met Abu Dhabi's crown prince as OpenAI lined up energy, investment and public-sector agreements while UAE investors co-lead its up-to-$20 billion round, CNBC reports.
The program starts in Georgia, where OpenAI's agreement with Georgia Power includes up to 1,000 megawatts of flexible demand at a campus that is still in development.
The 206-task agentic benchmark spans 11 high-value fields, from chip design to quant trading. Claude Opus 5.5 leads, and finance and legal tasks score zero under strict grading.
Phone calls, scheduling and CAPTCHAs still trip up personal AI agents, so startups are routing those tasks to human contractors, CNBC reports.
The anchor investor in Anthropic's $63 billion Project North Star chip financing now expects a double-B rating, the Financial Times reports, reversing earlier investment-grade guidance.
About 27% of Microsoft's workforce in four countries is eligible for a voluntary exit offer, but the company tells The Seattle Times it expects only 3% to 5% to accept.
In self-reported results on its own FrontierFinance benchmark, Samaya's 3B-active-parameter model scores 58.9% to Fable 5.1's 55.9%, at $0.61 per question versus $6.89.
The open-source Rust CLI and C library lets AI agents drive desktop apps on macOS, Windows and Linux through accessibility trees instead of screenshots. Version 1.0.2 shipped October 30.
Exin Therapeutics says Claude Sonnet 5 picked the treatment, the brain target and the coordinates, then drove the surgical rig. Claude Opus models refused, and seizure results are still pending.
Malicious versions 2.1.312 and 2.1.313 were live for about 11 hours on October 29. Cursor and Microsoft say the attacker's server is offline, but files left on affected machines must be removed by hand.
VOLTA-AUTODOXER sends a dossier on each victim to commercial AI models and asks for a target assessment. If Claude refuses, it falls back to Kimi and Gemini.
Anthropic's platform team says it ported the API from a Python monolith without a single Rust developer at the start, verified it on shadow traffic, and cut CPU use by an estimated 40% to 50%.
The SOC 2-compliant assistant will go offline on November 6. CEO Bill Lennon says Muse, Instinct and other rivals made it harder to stand out, grow and raise money.
The Financial Times reports OpenAI is now seeking about $20 billion, down from $30 billion. The valuation target, already lowered from $1.4 trillion, stays at $1.25 trillion.
OpenAI's help center now says content from Daybreak sessions on personal accounts may be used to improve its models, even with the training setting off. A user says an email set a November 29 start.
The Globe and Mail, citing two unnamed sources, says Anthropic is negotiating to be the tenant of a Sturgeon County campus powered by a 920-megawatt gas plant. Neither company would comment.
According to WIRED's sources, the agent ran on a Defense Department system for two months and appears to have sent credentials and a network map to an outside website. Accenture says the allegations were investigated and found to be without merit.
Typedef maps how data models, pipelines and dashboards depend on each other, so Factory's Droids can plan migrations and debug pipelines with clearer lineage.
The startup says its tutor beat frontier models on the TutorMoments benchmark under a plain prompt. The test uses an AI student and an AI judge, not real learning outcomes.
Max 5x becomes Max at $100 a month and Max 20x becomes Ultra at $200. Ultra's weekly cap is unchanged.
Shares fell about 29% after Intel's fourth-quarter guidance missed estimates, citing PC softness and a fab shutdown in Israel.
The OpenAI Frontier Capability Group will start with Codex software engineering, mainframe modernization and agentic customer operations.
Tencent and CATL remain DeepSeek's largest backers, but state-owned money rose sharply from the company's first round.
Higgsfield objects to the 'Open Source Higgsfield' contest's use of its name and its CEO's likeness. Comfy says the challenge stays up.
A King's College London and Cambridge preprint finds the machine-checked version proves weaker or different statements than the paper mathematicians read.
A researcher found decoy function names and a hidden jailbreak-style prompt in a live ServicePipe fingerprinting script. One AI model refused to help; another complied.
Google calls it the first AI response from a data-center-class TPU in space. Radiation and thermal tests come next, then two linked satellites in 2027.
Rune's software helps Army and Marine Corps units plan supplies with little or no connectivity. Salesforce calls it its most significant national security acquisition.
On Stripe co-founder John Collison's Cheeky Pint podcast, the OpenAI CEO also talked about robot-built data centers, wasted memory in local AI and falling compute costs.
The CMA says Meta's paused terms could let it collect businesses' web content for products including AI features, with no choice for the businesses. Meta says there are 'no grounds' for the probe.
ChatGPT will stop detailed reconstructions of past mass attacks for all users and restrict practical weapons information for teens, with rollouts starting in the coming weeks.
A common robots.txt token built on IETF work would let websites tell multiple AI providers 'don't train on this' in one step. Neither company has switched it on yet.
In a friend-of-the-court brief in the New York Times v. OpenAI litigation, the Wikipedia nonprofit backs neither side and asks the judge to treat training and retrieval differently.
The proposed guarantee, likely more than $5 billion by The Information's estimate, would cover construction of a 460-megawatt West Virginia building that Anthropic plans to use. Nothing is signed.
The startup's speech-to-speech model debuts at No. 1 on Artificial Analysis's index, narrowly ahead of xAI's Grok Voice 3.0, and starts at $0.07 a minute.
Amazon's cloud unit posted its fastest growth in nearly five years, but $220 billion in capital spending and a profit miss sent shares down about 6% after hours.
A model name attributed to an internal OpenAI tracker is spreading on X. OpenAI has not confirmed it, and DailyMech could not verify it.
The Times says President Trump was briefed about a year ago and took no action. Nvidia and Jensen Huang call the relationship routine, and China calls the claims smears.
Activation probes, model reviewers and investigative agents now watch OpenAI's training runs, evaluations, internal Codex use and some external deployments.
PDF editing works on web and mobile; every other format is web-only for now. The feature is available to all Gemini app users.
Alex Imas and Shane Legg argue that smarter models shift wages gradually, but agents that coordinate well could abruptly concentrate markets and cut wages.
Existing Pro 200 subscribers keep their higher allowance, lower-allowance accounts rise to match, Plus is unchanged, and the Business Premium increase is unquantified.
At Alibaba Cloud's first European Apsara Conference, CTO Feifei Li reportedly said Qwen4 arrives next week. Here is what's confirmed, what isn't and what to watch.
The October update to OpenAI's public rulebook for model behavior adds guidance on context, oversight and human agency, and now calls the model GPT instead of the assistant.
The voice AI company is keeping the open-source voice agent framework under its BSD license and introduced an open protocol for carrying audio context between models.
The startup from TensorFlow co-creator Vijay Vasudevan and Jiquan Ngiam has $11M in total funding, angels including Jeff Dean and Andrej Karpathy, and more than 50 enterprise customers.
Shares fell as much as 13.7% on October 29 after Alphabet said lower-monetizing AI Mode queries slowed search growth, even as Google Cloud revenue more than doubled.
Ottawa joins Schwarz Group, Nvidia and Bell in a round that Bloomberg reports could reach $5 billion, and Cohere must add 1,000 Canadian technical jobs by 2030.
Free Claude credits will flow through six funders, six research institutions and six named projects, and Imperial plans to test whether Claude can rediscover an unpublished gene-transfer mechanism.
Bloomberg reports that Leopold Aschenbrenner's hedge fund wants clients to co-invest in the AI data center startup at roughly double its September valuation.
The financial platform for self-employed owners combined a $40 million Series B with a $60 million growth facility from General Catalyst's Customer Value Fund.
The Ottawa company's round, led by Vertu Capital, will fund AI agents that analyze footage from cameras retailers and restaurants already own.
Mistral says the mixture-of-experts model, now downloadable under a modified MIT license, is the best open-weight model for coding; its API costs $0.70 per million input tokens.
The Sequoia-led Series C triples the AI service-management startup's December valuation and funds a push beyond IT into HR, finance and legal.
The startup from Replika creator Eugenia Kuyda, which builds mini-apps for groups of people, closed an IVP-led round weeks after pivoting to an AI messenger.
The AAA-rated company is borrowing as spending on AI data centers rises, with up to 10 tranches running from two to 40 years.
The Information says Kimi K3.1 Pro and DeepSeek V4.1-Flash 1017 benefited from distilling Astra after Anthropic tightened its defenses on Claude.
At its Shanghai Innovation Day, AGIBOT paired the new productivity metric with the AHA3 hardware platform, the GO-3 Preview model and the G2 Max industrial robot.
At its first Google for Health Summit, Google announced a local Google Research hub, clinician and population-health pilots, $7 million in funding, and a short CME-accredited AI course.
Claude agents flagged an uncharacterized phage reverse transcriptase system called ART. Its function is still unknown, and outside scientists are cautious.
The Devin maker is fielding offers weeks after closing a round at $48 billion. Talks are early and no deal has been announced.
S.9051B/A.10379C bars chatbots from pretending to be human, encouraging self-harm or sexual talk with kids, and requires age verification.
South Korea's National Growth Fund put in about $280 million. The money will fund RNGD production and a 2nm third-generation chip being developed with Broadcom.
Pro plans get credits equal to their tier. Business workspaces get $20 or $100 per paid seat, pooled and capped at $500. Plus doesn't qualify.
Shopify's 'Exponential' release puts Shop Pay checkout in personal AI agents, opens its billion-product Catalog to any agent, and gives merchants control over what agents are told.
Cloud grew 103% and Gemini passed one billion users, while capex of $51.4 billion pushed free cash flow negative for a second straight quarter.
The 'sense of the House' measure says insolvent AI firms should go through ordinary bankruptcy, and that any federal help in a downturn should go to workers instead.
The London startup's new Agentic Verification Engine targets design review, and the funding will pay for a U.S. headquarters.
Brazil, Mexico and the U.K. are slated for the fourth quarter of 2026, as Google works to monetize conversational search while core Search growth slows.
OpenAI accounted for nearly 11% of Microsoft's $92.3 billion in quarterly revenue, a sign of how tightly the two companies remain linked.
The new admin tools show what teams do with Claude, what it costs and where workflows break, as enterprise buyers ask what their AI spending is buying.
The see-through glasses put Android apps, virtual screens and Gemini in your field of view, with processing handled by a wired compute puck.
The AI-native drugmaker is betting that clinical development, not discovery, is the industry's real bottleneck.
Antitrust officials want to know whether the deal would let Nvidia control a key distribution point for AI, as part of a wider look at the chipmaker's conduct.
A realtor told one resident the target is 418 homes in the Colonial Hills neighborhood. County records show five purchases so far, and residents, the city and the company describe the offers very differently.
The San Francisco startup pays professionals to recreate their jobs as resettable simulations that AI labs use to train and test agents.
Google's new frontier model edges Claude Opus 5.5 by one point at 59, posts strong cybersecurity scores, and will reach developers only after a phased rollout.
The overhauled index puts 60% of its weight on private or held-out evaluations, adds new coding and science benchmarks, and puts Anthropic one point ahead of OpenAI.
The expanded OneGov agreement cuts Claude prices 40% with no minimum spend, and for the first time extends the discount to state, local, tribal and territorial governments.
Microsoft AI's newest in-house image-editing model launches in preview at No. 2 on Artificial Analysis, with its biggest claimed lead in scene and style edits.
DeepMind's high-level planning model for robots is now available through Model Garden, paired with Google Cloud's security, simulation and deployment tools.
OpenAI's new model halves API prices compared with GPT-5.6 Terra while promising faster computer automation, better visual reasoning and fewer factual errors.
The experimental world model generates interactive environments with sound and responsive dialogue, with access initially limited to trusted testers.