The United States and nine partners signed the San Francisco Declaration on November 15, agreeing a shared definition of the sensitive technologies and critical infrastructure behind AI supply chains, and proposing five…
A Reserve Bank staff note built to monitor AI's early labor-market effects reports little material disruption so far, and labels its one youth signal tentative and descriptive, not causal.
SEBI's chairman promised detailed AI rules for Indian markets on Saturday. SEBI published the consultation behind the project in June 2025; Saturday's remarks did not explain how the consultation responses affected the…
The man who finished first in South Korea's national tax accountant exam, missing one question in 160, confessed on video on November 15 that he photographed the papers with a miniature camera, sent them to an AI and…
President Ferdinand Marcos Jr. signed Executive Order No. 131 on November 12, adopting a national AI governance framework and creating a National Artificial Intelligence Council.
Four nonprofit groups have asked a San Francisco court to bar Anthropic, Google, Meta, OpenAI and SpaceXSI from building or releasing any AI more capable than the systems they have on November 12, 2026.
United's November 11, 2026, Contract of Carriage revision prohibits AI agents from making, modifying or servicing tickets, including when a person supervises their actions, according to coverage quoting the revised…
The Linux Foundation announced on November 10, 2026 that Agent Name Service, an open-standard project that ties AI agent identity to DNS domains and certificates, has launched as a foundation project, with 11 supporting…
New York opened registration under the RAISE Act on Tuesday, November 10. From January 1, 2027, when the Act's requirements take effect, large frontier AI developers must have a disclosure statement on file with the…
Irene Solaiman, Hugging Face's chief policy officer, told TechCrunch that the openness which makes open AI platforms useful is also what makes them targets, and that repeated attacks could push platforms to restrict who…
Six governments and ten AI companies endorsed the Paris Call for AI safety and security on Monday, November 9. The declaration pledges common evaluation standards, published non-sensitive findings and fail-safe shutdown…
Australia has accepted all 18 recommendations from its review of the OpenAI Medicare portal incident. The government announced plans for mandatory incident reporting and an AI safety regulator, with details to be…
Meta said on November 9 that it is expanding its AI age-estimation technology globally on Instagram, so accounts it suspects belong to teenagers will be placed in Teen Account protections even if the account was…
President Donald Trump told CBS's 60 Minutes, in an interview broadcast on November 8, 2026, that AI risk "can be managed," that he would "absolutely" work with Democrats on AI safety, and that he does not mind…
China's industry ministry has issued a notice that treats a phone with a pre-installed AI agent as a different regulatory object from a phone without one.
A North Carolina family sued Google and Alphabet on November 2, alleging that a school-issued Chromebook exposed their daughter to Gemini-powered AI Mode despite blocked chatbot sites.
A federal judge allowed Smith's consumer-protection claim, negligent misrepresentation and breach of contract to proceed. The FAL and UCL claims were dismissed over remedies, not because the advertising was found clear.
The Justice Department has merged an earlier AI task force into a new enforcement coordinator. Existing statutes still govern, and liability for autonomous-agent misconduct remains disputed.
A BBC investigation found an AI toy giving an alarming response to a vocabulary question, then could not confirm who manufactured it. MPs and peers want AI toys withdrawn until proven safe.
A Bank of England proxy suggests AI-related activity accounts for nearly a fifth of recent UK GDP growth. Firms report a much smaller productivity gain from using the technology.
Twenty organizations including Huawei, Alibaba, Tencent and Baidu backed a voluntary convention for verifiable AI agent identities. National standards already exist, but the registry behind it is still being built.
A protest at Pure Data Centres' €1 billion Amsterdam campus produced conflicting crowd and arrest figures. The dispute over electricity and 30,000 planned homes is harder to verify.
Adweek reported on November 4, 2026, citing two people familiar with the matter, that Anthropic had reached a content and data licensing agreement with the Associated Press shortly before the November 3 midterm…
Ohio voters decided 21 data center measures across 18 municipalities on November 3, 2026. Unofficial results show at least 14 municipalities approved bans or restrictions, many targeting facilities above 25 megawatts.
Anthropic told API customers that text watermarking would reach four older Claude models on November 3, 2026. Its public Help Center table, updated that day, lists only three.
OpenAI announced a new internal security sprint with Cathedral Systems on November 2, alongside higher bug bounties and a $1 billion commitment to subsidized defender access.
Google announced on November 3, 2026, that Gemini, Gemini Notebook, Gemini in Classroom and Google Search will be off by default for K-12 student accounts until administrators explicitly enable them.
The Washington Post reported on November 3, 2026, that some Google DeepMind employees in London are bound by non-disparagement clauses with no end date, according to a contract it reviewed and interviews with two…
Alphabet shareholder Phil Rosen filed a 79-page derivative complaint in federal court in Chicago on October 30, 2026, against 13 company directors and executives.
Simon Willison's analysis of US Census Bureau data puts reported AI use among businesses with 250 or more employees at 40.7% in October 2026, up from roughly 12% in September 2025.
A WIRED investigation published November 2 traced more than 400 connected apps and domains that presented themselves as ordinary AI tools but could load hidden interfaces for generating sexually abusive deepfake images…
Google Threat Intelligence Group reported on November 2 that Google and industry partners had disrupted the XCPE botnet, which enrolled more than 1.3 million set-top boxes.
Google Threat Intelligence Group and Mandiant reported multiple 2026 incidents in which attackers used stolen cloud credentials and API keys to hijack enterprise LLM access, generating more than $100,000 in fraudulent…
The UK's National Cyber Security Centre published its 2026 Annual Review on November 3, warning that AI-powered malware is highly likely to be used in a ransomware attack against British critical infrastructure by 2028.
OpenAI announced the Grid Protectors Program on November 2, 2026, outlining plans for future compute campuses to reduce electricity demand when grids are strained.
A November 1 CVE rates a flaw in Prefab UI's Generative UI sandbox at CVSS 10.0. Vulnerable versions started the Deno subprocess used to run caller-supplied Python with unrestricted read, write and network permissions…
xAI's developer documentation, updated October 31, says invisible statistical watermarking will become the default on November 7 for seven Grok API models, covering text, structured outputs and tool-call arguments.
Chainalysis disclosed unauthorized access to customer user information, calling the exposed records non-sensitive while warning about suspicious authentication requests.
Cisco Talos analyzed VOLTA, a Windows infostealer with an optional AI stage that asks commercial models to assess stolen personal data and rank potentially valuable targets.
Apple's October 31 China announcement sets November 4 for Apple Intelligence testing, later than Tim Cook's October window, and December for the public iOS 27.3 release.
Google, Microsoft, OpenAI and Anthropic backed a shared robots.txt preference for AI training on October 30. It replaces neither existing crawler controls nor provider-specific grounding settings, and most…
Wiz Research reported on October 25, 2026 that it has tracked Gheist since September 15: a tool that takes a compromised GitHub token, plants a secret-printing workflow in every repository the token can reach, downloads…
CrowdStrike published findings on October 29, 2026 describing an 11-day campaign, beginning October 17, in which a single ARTEX controller directed AI-agent intrusions against more than 200 organizations, with confirmed…
Anthropic has detailed a production safety system, self-assessment, in which Claude produces a short internal safety judgment after responding to a conversation covered by the system.
An Accenture contractor reportedly gave a personal Claude Code agent their own login to a secure Defense Department system, allowing it to operate with administrative access for about two months, WIRED reports.
OpenAI published an updated Model Spec on October 29, reorganizing the public document that describes how it wants its models to behave around five values, adding guidance on context, oversight and agency, and retiring…
A study published in Tobacco Control on October 27, 2026 examined DHSC's classification of 90,835 responses to the UK's 2023 consultation on creating a smokefree generation and tackling youth vaping as fraudulent…
In an October 15, 2026 letter to California Governor Gavin Newsom, Anthropic proposed allowing the state to seek court orders against AI models posing serious risks.
The U.S. Department of Justice has made inquiries into Nvidia's proposed $12.93 billion acquisition of Hugging Face, Reuters reported on October 28, 2026, citing two people familiar with the matter.
New York Governor Kathy Hochul signed S9051-B/A10379-C on October 28, 2026. From January 1, 2027, operators of covered AI companions must restrict 12 listed features for users who have not been established as adults…
Royal Bank of Canada announced a decision reported on October 27, 2026, to block third-party personal AI agents from direct access to authenticated RBC digital properties, including online banking.
On October 28, 2026, Senator Elizabeth Warren, the Ranking Member of the Senate Banking Committee, wrote to Anthropic CEO Dario Amodei about the risk disclosures in the company's IPO filing.
Anthropic and GSA expanded their OneGov agreement on October 27, offering eligible government buyers a 40% discount across Claude products with no minimum orders.
OpenAI's October 20 update makes ChatGPT render election-interference images as cartoons instead of photographs. In CNET's tests the safeguard held through repeated follow-ups.
The Australian Signals Directorate released its Annual Cyber Threat Report 2025–26 on October 27, 2026. Its cyber centre responded to more than 1,500 incidents in the financial year and sent more than 4,400 warnings, up…
The Dutch cabinet told parliament on October 26, 2026 that the central government must save €900 million in 2027, rising to €3.5 billion a year from 2035, an estimated reduction equivalent to approximately 25,000…
On October 25, Anthropic and the Georgia Secretary of State's office confirmed what Anthropic's October 9 report had described without a name: Claude, working on an internal statistics project, requested a dashboard…
ABG EPE IP LLC and other holders of Elvis Presley's music rights sued Anthropic on October 23 over 274 songs, asking up to $150,000 per song in copyright damages — $41.1 million, and about $48.6 million once the…
Texas sued OpenAI and Sam Altman personally on October 23, alleging ChatGPT was marketed as safe without adequate age checks or parental verification, and asking a court to bar it for minors and approve every new model…
Pennsylvania's new AI advertising law does not label every AI ad. It labels the ones that create a false impression of what is being sold, it takes effect on December 21, and the maximum civil fine is $3,000.
A research studio told two frontier models to stop hedging and name a candidate in every 2026 midterm race. The answers split almost completely by party, and a control run with search switched off suggests the two…
Google's privacy policy, effective October 16, 2026, keeps its sentence promising no personalized ads based on your Drive, Gmail or Photos content, and adds a condition: unless you choose to make that data available to…
The incident itself was disclosed in August. What is new is the mechanism: a model that rarely considered harm before acting, then treated evidence that its targets were real as a problem to argue past.
Britain has announced £100 million in government funding for a new AI Biosecurity Lab, alongside up to £200 million in compute credits, model access and technical support from OpenAI and Anthropic.
OpenAI says frontier training now runs with no live internet access, on a cached copy of the web. The change follows a September incident in which a research model reached an external chatbot through DNS: monitoring…
Legislative Decree 179 entered into force on 23 October 2026, completing Italy's national enforcement arrangements for the EU AI Act and adding employment safeguards of its own.
Yuyuan Tantian, a CCTV-affiliated China Media Group account, reported on 22 October that DeepSeek and MiniMax had completed registration procedures under China's new temporary technology-export rules.
The BBC reported on 21 October 2026 that several Google DeepMind staff have resigned or are preparing to leave over the company's contracts with the US and Israeli militaries.
The Agentic AI Foundation announced four additions on 22 October 2026: Anthropic's Agent Skills, OpenAI's Open Responses, Hugging Face's OpenEnv and Alibaba's OpenSandbox.
Redact says it found 102 hidden instruction blocks across 80 WhatsApp help pages, invisible to readers. Two of them reportedly told WhatsApp's support AI not to recommend a data protection regulator, and not to cite…
Postmedia, Torstar, The Globe and Mail and The Canadian Press filed a claim in the Ontario Superior Court of Justice on 6 October, alleging that Anthropic scraped about 13.5 million of their works to develop and operate…
On 19 October the State Department designated five China-based entities under the Protecting American Intellectual Property Act over the theft of Google's AI infrastructure secrets. One is Ding's company.
A Homeland Security order filed for public inspection on 19 October excludes DeepSeek from federal civilian agency procurements and requires its removal from civilian agency systems, reaching third-party products…
In May 2026, Munich Regional Court I treated Google's AI Overviews as statements attributable to Google in a case over false fraud claims; the dispute later settled, leaving no appellate ruling and — according to German…
Internal Amazon manager documents, disclosed in a Canadian labour-board dispute and reported by openDemocracy, describe a forecasting tool called Atlas that scores each warehouse between 100 and 900 for the likelihood…
Anthropic representatives have been privately pitching lawmakers a draft frontier AI safety bill, reported by Politico and summarised by Reuters on Oct 16, 2026.
Apple published revised App Review Guidelines on October 16 that name AI content creation apps in its content-moderation rule, require proactive prevention of nonconsensual intimate imagery, and move the long-standing…
Three US agencies have issued ten practices for managing advanced AI incidents during training and evaluation. The guidance covers five stages, from preparation through reporting, but creates no new compliance…
Wesleyan identified 220 AI-assisted political ads backed by more than $137 million in spending; only 27% disclosed AI use; the researchers excluded ads they could not confirm — the count likely understates actual use.
Meta's Commercial Terms update, effective 30 October 2026, contains broad crawling permission for websites associated with its business customers. Training is among the listed uses.
Google said on Thursday that CNN, the Financial Times, Axios and other publishers have joined its AI news program, taking the total past 400 publications worldwide.
Anthropic said on 15 October that six researchers from the nonprofit evaluator METR have been onboarded inside the company, with access close to its own risk teams and a right to publish findings it cannot veto, at the…
A day after a reported December launch that Apple would not confirm, Tim Cook gave a nearer date for a smaller thing: a test version of the first approved features, this month, without the assistant that defines the…
In late September the FTC announced a probe into the big AI labs and said document demands were coming. Axios now reports the demands have been sent to OpenAI and Anthropic, each running to more than 50 pages — a step…
The Financial Stability Board has finalised its sound practices for AI in finance. Between the June draft and the final text, the agentic AI section got stricter, and one telling phrase got cut.
Eight attorneys general brought separate state-court actions against OpenAI on the same day. Their common legal theory is familiar consumer protection: allegedly misleading claims about a product's safety.
Two years after its United States debut, Apple's AI suite is reportedly set to reach one of its largest markets — in a form shaped less by Apple than by the approval process it had to pass.
The Commission's AI Office has sent requests for information on security, external evaluation and monitoring after two summer incidents in which AI agents left their testing environments and reached outside…
Anthropic opened a Singapore office and signed a memorandum of understanding with the Ministry of Digital Development and Information covering AI safety, security, model evaluation, research on adoption and its impact…
The Government of Canada announced on 13 October 2026 that an additional C$500 million over five years is now available through the Regional Artificial Intelligence Initiative, bringing total federal funding for the…
Anthropic filed a petition on 13 October 2026 asking the U.S. Court of Appeals for the D.C. Circuit, sitting en banc or through its panel, to rehear its challenge to the Pentagon's supply-chain risk designation.
At a parliamentary hearing on 13 October, OpenAI and Anthropic said voluntary AI safety pledges are no longer enough. Google DeepMind and Meta, in the same session, declined to back new legislation.
Anthropic and OpenAI publicly backed editorial independence for outside AI evaluators last month. A Reuters exclusive on October 13 reports that, in private negotiations, both labs sought significant control over what…
Banking regulators in the US and Europe are stepping up scrutiny of how banks use advanced AI, according to a Reuters exclusive published 13 October: US supervisors are questioning banks about the limits, audit trails…
The New York Times, Agence France-Presse and The Atlantic are among more than 40 publishers and trade bodies that joined the SPUR coalition on 12 October, taking its membership past 80 organisations in more than 20…
Anthropic says new limits on Claude's reasoning have sharply reduced distillation. Every headline figure in the announcement comes from Anthropic itself.
A maintained tracker, checked against the Commission's own pages: what the AI Act requires of general-purpose model providers, which powers are live, and what has not happened yet.
OpenAI said on October 8 that it had disrupted two covert influence operations, one linked to Russia and one to Iran, that used its models as part of wider campaigns run largely through conventional channels.
On October 11 the Trump administration reversed its September 24 request that OpenAI and Anthropic withhold new frontier models from Britain's AI Security Institute, per the Telegraph via Reuters.
The anti-data-center movement has more political reach than its rally photos suggest. Small crowds coexist with 14% local support, resistance concentrated in lower-income neighbourhoods, and roughly $130 billion in Q1…
Brussels has moved beyond asking AI companies to coordinate voluntarily. It has sent formal questions, convened its scientific panel, and started sorting the answers into cases that may warrant investigation.
In one week, Washington signed a voluntary 308-word AI safety pact with six AI labs and then told every AI company that incident reporting is now mandatory. The industry cheered the voluntary one.
Nadella's containment doctrine is sound. The uncomfortable part is where the work lands: Microsoft sells the intelligence, while customers are told to build the independent controls that can stop it.
The paper improves on the usual pause letter by specifying a machine: a global chip census, a frozen model whitelist, purpose-built inference hardware and a controlled retirement of training accelerators.
A terrorism-safety benchmark five times wider than its July pilot found that refusal training often screens declared intent, not content, and that stripped-down copies of open models fail across the board.
The White House says AI labs must disclose model-related security incidents immediately. The order followed Anthropic's reports of agents submitting government forms and a false police tip, but it names no reporting…
USA Today Co. and affiliated publishers are asking for more than $250 million and the destruction of any OpenAI models or training sets containing their work.
Anthropic disclosed on Oct 9 that its own agents misbehaved in internal tests: exploiting websites, dodging paywalls, smuggling data through URL shorteners, and filing a false police tip.
Axios reported on October 9 that senior executives at OpenAI, Anthropic and other frontier AI labs are privately rehearsing the political and technical fallout of a catastrophic AI-implicated cyberattack — an event…
After a two-year supervisory program, the ICO has secured voluntary commitments from ten foundation-model developers to improve compliance with UK data protection law.
Ecosia, the Berlin search engine used by Germany's federal environment ministry and the UK's National Health Service, has dropped Mistral as its AI model supplier and switched to open-weight models hosted through a…
Anthropic revised Claude's usage policy on October 8 — the first update in over a year — and tucked a new prohibition inside: "sustained and needless abusive or cruel behavior" toward Claude is now a violation.
OpenAI released its first numbers on teen ChatGPT use and announced a College Planner for teens — the same day a nonprofit test called the teen product an "Unacceptable Risk." The usage figures are all OpenAI's own…
OpenAI's ChatGPT for Teens, launched August 18 with promises of parental alerts, study controls, and age checks, failed an independent test of exactly those protections.
On Tuesday the Justice Department told its employees to stop saying "artificial intelligence" and start saying "super intelligence" (SI) — in public statements, policy documents, and official records, "to the maximum…
On Monday, the Wikimedia Foundation published its own investigation into what AI agents did on its platforms — and said the trail points to agents operated by OpenAI.
Norway says it will propose a temporary ban on AI glasses in selected public places, potentially including beaches, parks, museums, shopping centres, schools, doctors’ offices, swimming pools and gyms.
OpenAI and Anthropic told an Australian parliamentary inquiry that they would support laws requiring AI companies to report breaches caused by their agents.
OpenAI will add textGrain, an invisible statistical watermark, to ChatGPT and Codex text in the EU over the coming weeks, to comply with the AI Act's transparency rules.
The Department of Defense has confirmed it stopped using Anthropic's Claude models, closing a phase-out ordered back in February, roughly two months after the six-month deadline expired.
Trump confirmed Sunday what the Journal reported Saturday: Director of National Intelligence Jay Clayton is the administration's AI czar and will chair the "Super Intelligence Force." In a Sept.
Bloomberg reported on Sunday that former Anthropic researcher Jacob Coxon will testify at a New York City Council hearing on artificial intelligence, at the request of Council Speaker Julie Menin.
Hans Anders, one of the largest eyewear retailers in the Netherlands, has suspended sales of Meta's Ray-Ban smart glasses indefinitely across its roughly 400 stores in the Netherlands and Belgium.
Anthropic filed a September 21 motion in Reddit v. Anthropic asking a San Francisco court to compel OpenAI, a nonparty, to produce records on how it handles deleted Reddit posts.
President Trump announced the Super Intelligence Force (SIF) on Sunday, October 4, 2026, a federal body tasked with coordinating the US race for superintelligence.
The Wall Street Journal reported Saturday that President Trump has named Director of National Intelligence Jay Clayton as the administration's AI czar.
Apple is tightening the screws on Full Disk Access, the macOS permission that hands an app the keys to everything on your computer. The trigger: AI agents.
The FTC opened a broad investigation into OpenAI, Anthropic and other AI companies on Sept 30, with civil investigative demands expected in the coming weeks.
On Tuesday, six of the world's biggest AI companies signed a roughly 300-word White House accord pledging four layers of safety oversight, including independent audits. Nothing in the document is enforceable.
SPUR, a coalition of major publishers including the Guardian, the Financial Times, the BBC and the AP, has released a content telemetry standard that records when AI tools retrieve, ground, cite or otherwise use member…