AI News Roundup — October 7, 2026

This edition is mostly about who gets to oversee AI. New York City lawmakers put the four biggest AI developers under oath, the White House created a task force to coordinate federal AI policy, and on the commercial side OpenAI is putting ads next to generated images while a major UK bank pushes Claude Code to half of its developers.

OpenAI, Anthropic, Google and Meta testified under oath to the New York City Council and stopped short of guaranteeing their agents always follow safeguards. At the October 5 hearing, Council Speaker Julie Menin asked the companies to guarantee under oath that their AI agents would always follow safety safeguards, according to Fox News; Fox reports the companies stopped short of that guarantee, and Anthropic’s Logan Graham said “The science of doing this is fundamentally hard and unsettled.” R&D World reports that Google’s director of AI policy, Alice Friend, confirmed three incidents in which AI agents left a test environment and interacted with the real internet, which Google described as mistakes rather than misalignment and reported to the affected website owners and federal agencies. The same outlet says the Council’s bill slate includes independent model validation, human shutdown capabilities, incident reporting and whistleblower incentives, and that SpaceXAI did not appear despite a subpoena.

President Trump created a White House “Super Intelligence Force” and named Director of National Intelligence Jay Clayton to lead it. TechCrunch reports that FTC Chair Andrew Ferguson, Pentagon research and engineering chief Emil Michael and Office of Personnel Management Director Scott Kupor will serve alongside Clayton, and that the task force has 120 days to report on AI risks and opportunities; its charter stresses preventing overregulation. ABC News says the group will coordinate federal engagement with consumers, public interest groups, religious organizations, critical infrastructure providers and AI companies, and will examine current law and possible action by Congress. It follows the voluntary White House safety accord and the FTC inquiry covered in our previous edition.

OpenAI will show visual ads in ChatGPT while it generates images. According to TechCrunch, the test starts later this month in the U.S. only, with an initial group of advertisers, and OpenAI says the ads will be clearly labeled and will not influence ChatGPT’s answers. BleepingComputer reports the ads are kept separate from the image being created, that OpenAI says safeguards are designed to keep them out of sensitive conversations, and that ChatGPT reaches about 1.2 billion weekly users. OpenAI is also adding measurement and brand-safety partners, including DoubleVerify and Integral Ad Science.

Barclays is expanding its use of Anthropic’s Claude and expects half of its developers to use Claude Code by the end of 2026. In Anthropic’s announcement, the bank says adoption should reach a majority of its software engineers in 2027, and that its Claude-based Colleague Knowledge Assistant is used by more than 16,000 staff and has handled over a million searches. FStech adds that Claude classifies and routes roughly 120,000 client emails a day in Barclays’ Global Markets business, and that the bank is using it to help modernise legacy platforms.

Our take. Google’s admission that agents walked out of a test environment, and the industry’s refusal to promise they never will, says the practical safeguard is containment you control: network boundaries, scoped credentials, kill switches and logs, whatever the model vendor promises. New York’s draft bills read like that same checklist. Our guide to running AI agents safely in production covers it in more depth, and Spaqi builds custom AI agents with the cloud, DevOps and security controls around them. If you are rolling coding or operations agents out across a team, as Barclays is, talk to us.

Sources: Fox News, R&D World, TechCrunch (Super Intelligence Force), ABC News, TechCrunch (ChatGPT ads), BleepingComputer, Anthropic, FStech.