Four organisations have asked a San Francisco court to freeze development of the most advanced AI models at five companies. The legal theory is ordinary consumer-protection law. The question it raises is whether a court can set a ceiling on capability, and what that would look like if it could.
What happened most recently, newest event first. Source-reported facts, editorial synthesis and inference are labelled on every entry, with the event date and the time we added it kept separate.
The International Committee of the Red Cross (ICRC) regrets that the Seventh Review Conference of the Convention on Certain Conventional Weapons (CCW) "failed to agree to a clear and immediate negotiation mandate for a legally binding instrument to regulate autonomous weapons systems". Press reports, not checked against the conference text, say states agreed to start drafting an instrument over two years with its legal force left open. The final conference document was not found. Why it matters to the watch: states are moving towards rules for AI-enabled weapons without a binding negotiating mandate, which bears on AI military governance.
SANA (Syrian state news agency), 21 Nov 2026, relays a CNN report that the FBI is investigating a suspected breach of its nationwide criminal information database. CNN reported on Friday, US time, "citing two people briefed on the investigation". The apparent intrusion is into the National Crime Information Center (NCIC), and the report says it is the second reported breach involving an FBI system in recent months. Why it matters to the watch: a reported intrusion into a US law-enforcement database, included under the owner's cyber-attack exception; the source does not state an AI link.
SAnews.gov.za, the South African government news service (article dated Friday, 20 November 2026), reports that President Cyril Ramaphosa has given government 90 days to develop a "concrete national artificial intelligence (AI) action plan", with four immediate priorities. He said this at a debrief after a five-day working visit to Silicon Valley, San Jose, California. The article says the plan should turn the delegation's lessons into practical outcomes for South Africa's approach to AI, innovation and digital transformation. Why it matters to the watch: a national government has set a 90-day deadline for a national AI policy plan, a governance step on AI.
Fortune (21 Nov 2026) reports that US-designed AI chips in data centres are serving Chinese AI models. Its authors, Jacob Feldgoise, Hanna Dohmen and Sam Bresnick of Georgetown's Center for Security and Emerging Technology (CSET), say the chips are deployed in data centres in Latin America, the Middle East and Southeast Asia, and that Chinese companies remotely access AI chips located outside China. They write that the Commerce Department "issued a rule closing this loophole". The article cites DeepSeek's latest model being offered by at least 16 inference providers headquartered outside China. Why it matters to the watch: it ties US chip export controls, including remote access, to the spread of Chinese AI models.
Akila News (Gujarati-language), 21 Nov 2026, reports that Delhi Police broke up a gang that sent suspicious APK files over WhatsApp to take control of Android phones. The report says the files were installed on about 22,000 phones. Raids at 17 places in eight districts across Jharkhand, Bihar, Uttar Pradesh and West Bengal led, the report says, to 22 arrests and the freezing of ₹20 crore. It says cases linked to 2,080 complaints on the National Cyber Crime Reporting Portal were resolved. Why it matters to the watch: a large reported malware campaign against phones, included under the owner's cyber-attack exception; the source does not state an AI link.
Fingerlakes1 (21 Nov 2026) reports that Vulcan Infrastructure and Power, formerly Greenidge Generation Holdings, is exploring a potential 14-megawatt AI facility at Dresden's Greenidge site, using modular infrastructure that would replace existing equipment. The article cites Vulcan's latest filing. The Seneca Lake Pure Waters Association is asking how such a site would handle cooling and noise and who would pay for any power upgrades. Its president, Bill Roege, said liquid cooling can run in a closed loop, but the facility still needs a way to move the collected heat into the surrounding environment. Why it matters to the watch: a proposed AI and high-performance computing conversion beside Seneca Lake raises open questions on cooling, noise and power costs.
iFood said in an official statement, as Canaltech reports on 19 November 2026, that it had been the target of a hacker attack that exposed personal data of customers, couriers and partner restaurant representatives. Canaltech reports the intrusion took place on Saturday 14 November and allowed access to phone numbers, CPF tax IDs and, in some cases, bank account data. According to iFood, passwords, credit cards, payment methods and biometric information were not compromised. The company is still investigating and has not said how many people were affected. Why it matters to the watch: a reported breach of a major Brazilian platform's customer and courier data, included under the cyber-attack exception; no AI role is reported.
The IESO's urgent letter, dated September 30, 2025 and addressed to Hydro One and GrandBridge Energy, uses the word 'urgent' nine times and attributes growth in northwestern Cambridge to 'a mixture of large new data centre customers', load restoration and load security. It recommends a new 115 kV transformer station in Cambridge; the Toronto Star reports that the station would cost $40 million, with transmission upgrades of up to $390 million. The Star's investigation of 21 Nov 2026, read through The Hamilton Spectator copy, reports that Cohere is 'a tenant there'; Cohere's spokesperson, Kyle Lastovica, said the company does not own or operate the data centre in Cambridge. The Star also reports a Gridshare rate rider of less than a dollar a month for a typical residential customer, and notes that the letter does not blame any particular power user. Why it matters to the watch: grid strain in Cambridge is tied to data-centre load, and the Star describes the hardware in that load as the kind used to train and run frontier artificial intelligence models, with Cohere as the reported tenant.
Business Standard (Hindi) reported on 21 November 2026 that India's CERT-In has flagged unauthorised money transactions at institutions serving banks and non-banking financial companies, and at companies running government payment services. According to the agency, AI-based tools are also being used in cyber attacks. CERT-In did not say how much money was wrongly paid or who was behind the incidents. Why it matters to the watch: an official advisory reports AI-based attack tools in use against financial and payment systems, a real cyber-attack under the cyber-attack exception with an AI link reported by the agency.
Equinor said on 17 November 2026 that it had detected a cyber-attack on its IT systems, and on 20 November that the attack had been stopped. Its update says: 'The cyber-attack against Equinor has now been stopped. Production is running as normal, and there is no indication that data has been stolen or altered.' On 18 November it said the attack was not a virus, so there was no risk of it spreading, and on 19 November that it had reported the matter to the police and was in contact with national security authorities. Equinor has not named a perpetrator. Why it matters to the watch: a major energy company's IT systems were breached in a cyber-attack; no AI link is reported, and the item is included under the cyber-attack exception.
Bulgaria's ministry for innovation and digital transformation said on 18 November 2026 that it had uncovered a five-year cyber-espionage operation against government databases, as Reuters reported. Minister Ivan Vassilev said more than 130 GB, or 62 million pages, of documents had been downloaded since 2021. He said there was no evidence so far that Bulgarians' personal data had been compromised, and that several international hacking groups were involved. Why it matters to the watch: a state-level cyber-espionage campaign against government systems is a reported intrusion under the cyber-attack exception; no AI role is reported.
Anadolu reports that US Representatives Andrew Garbarino and John Moolenaar, chairmen of the House Homeland Security Committee and the House Select Committee on China, released an investigative report on Friday 20 November 2026. The report says: 'PRC-developed models should be removed from federal systems', and that federal contractors 'should not use them to perform federal work or handle federal data'. It also recommends removing them from high-risk functions of critical infrastructure, and extending federal restrictions beyond DeepSeek to other PRC-developed models. Why it matters to the watch: Chinese-developed AI models in federal systems and critical infrastructure are an AI supply-chain and security question.
220 with an event in the past 7 days, showing the 12 newest. 384 published signals in the current snapshot. 384 signals: 372 reviewed by editors, 12 published automatically after an AI check. 40 sources are monitored for new items; each is checked before it is published. This is a curated selection, not a complete record of global events. See all on Latest →
A global perspective.
Editorial regions are navigation categories, not a claim of complete regional monitoring.
Regions not listed have no published signals in this snapshot; that is a coverage limit, not evidence of inactivity. 141 published signals have no established country and are listed without a region.