Agents not welcome
United Airlines is banning AI agents from booking flights for travelers. Its reasons: "sharing account credentials with personal, third-party, AI assistants can expose sensitive information, and actions taken by these tools may result in unintended purchases, account changes, or use of MileagePlus miles." It also cites federal rules, such as travelers confirming in person that their bags hold no banned items. Another blocking, after The RealReal in No. 4.
https://www.cbsnews.com/news/united-airlines-bans-ai-agents-booking-flights/
A reader told The Verge that his Meta Muse agent, which had collected watch listings from eBay for weeks, was suddenly blocked. The Verge could reproduce it, and sometimes Muse sent people to sites that mirror eBay's listings instead. eBay wouldn't say directly whether it had added new limits, only that it has "restricted agent access in certain cases." Its user agreement already bans "buy-for-me agents." Now eBay is also walling off its site to agents. Expect more to follow.
https://www.theverge.com/tech/1030838/ebay-ai-agents-meta-muse-agentic-shopping
Some jurors of the Independent Games Festival say they won't vote for games that use generative AI "in any capacity," after "a significant increase" in such entries. The first test case is Big Walk by House House, the makers of Untitled Goose Game. They say chatbot snippets were a starting point for "well under 0.1%" of the game's code, mostly to fix problems with third-party code, and that they disclosed it in their submission. The backlash is still growing.
https://www.theverge.com/games/1030994/big-walk-independent-games-festival-igf-ai
But there are also other kinds of agents. Nvidia's CEO Jensen Huang admitted to an affair with a woman U.S. intelligence believes is a Chinese spy, which Beijing calls "baseless." Nvidia's board opened an investigation. Then The New York Times reported a second affair, with an Nvidia executive, and The Wall Street Journal reported that the board is preparing to force him out. Looks like Huang won't be welcome at Nvidia much longer, after violating the ToS for agent use.
https://www.cnbc.com/2026/11/11/nvidia-jensen-huang-affair-employee.html
The tools change hands
At Adobe MAX, Adobe previewed integrations that let the AI agents you already use call Photoshop, Illustrator, Premiere and After Effects, with a private beta in early 2027. It also showed "always-on" agents that keep doing recurring work in the background, an AI-first editor in Photoshop with a switch back to the Pro Editor, and Firefly Builder, which turns an idea into a working app or website prototype. Too little, too late: Adobe is going to go away, like the dinosaurs.
https://www.theverge.com/tech/1030593/adobe-max-ai-agents-mcp-firefly-assistant-updates
Meta open-sourced Astryx, a design system that grew inside Meta over eight years and powers more than 13,000 of its apps: 150+ accessible components, seven themes, MIT license, still in beta. It calls itself "built for people and agents," because "the API, docs, and CLI are designed together so a person and an AI assistant build the same way, from the same reference." Interesting. Not sure where the "built for agents" part comes from beyond that, but I would like to know more.
https://github.com/meta/astryx
The Chrome, Edge, Firefox and MDN teams published Modern Web Guidance, skills that steer coding agents away from legacy patterns and toward the modern web platform, across about 250 web features, with accessibility built in. Good idea.
https://github.com/GoogleChrome/modern-web-guidance
Google released the Gemini Decisions API. Instead of writing text, its model scores a fixed set of options, like yes or no, a tool to pick or a policy check, and returns probabilities and confidence scores in 150 to 250 milliseconds. Interesting, decision models everywhere now.
https://blog.google/innovation-and-ai/technology/developers-tools/gemini-decisions-api/
Who checks the work?
Grok Build, the vibe-coding agent from SpaceXSI, gives each app a randomized share link. But a design decision made those apps easy to list anyway: security firm Claroc found about 35,000 of them open to anyone who looked, with internal business plans, medical exam results and employee ID numbers. Two dozen were built by SpaceXSI itself. This is not good.
VulnCheck checked on Anthropic's Project Glasswing, its AI hunt for security holes, seven months in. Of 37,966 reported findings, 217 are marked fixed. Since September, nearly 12,000 findings were added and 15 more fixes recorded. At curl, maintainers rejected 19 of 34 published reports as not being vulnerabilities at all. Wow, so many still to fix. Didn't expect progress here to be so slow.
https://www.vulncheck.com/blog/anthropic-glasswing-november-update
Pietro Miotti had 48 AI agents look through about 1,200 square kilometers of lidar images and found what looked like a Maya site that wasn't in Mexico's official register. "This is by far the best time ever to be alive." Then came the update: archaeologists had already mapped the site in a 2025 paper, which the agents missed in their literature search. Haha :) ...
https://pietromiotti.com/latest/maya-site
Also worth a look
Ring's flying camera
Amazon's Ring Drone Cam is a $249.99 indoor camera that flies routes through your home. Its AI maps the space and suggests patrol points, using vision models and cloud processing. It ships in the US in January, though it hasn't yet been authorized by the FCC. Amazon itself says "Ring Drone's propellers aren't quiet," so at least you'll hear it coming. Ok, this is creepy.
https://www.aboutamazon.com/news/devices/ring-drone-cam-rover-cam-security-robot
Palantir and your GP
WhatsApp messages obtained by the campaign group Foxglove show Palantir's UK boss pressing the then health minister to bring GP records into its AI platform, arguing that "the logic of joining up primary and secondary [care] data is unarguable." GP data is expressly excluded from Palantir's £330m NHS contract. Not a good look for Palantir, but I guess they don't really care.
Qantas cuts 850 jobs
Qantas will cut 850 jobs in Australia and move work to India under an agreement with Accenture, citing "unprecedented" technological disruption. It's the third big layoff plan in Australia in three weeks, after the design company Canva and National Australia Bank. Things are moving because of AI.
"Get ready to go to jail"
On 60 Minutes, Trump said someone was reported for "doing something bad with AI," and his administration told them: "Get ready to go to jail. Please give your wife a kiss good night." He wouldn't say who, only that it was the leader of "one of the big ones," who then "totally changed everything." I guess he is talking about Dario Amodei here. Trump singled him out on Truth Social in September, though Gizmodo notes that Amodei doesn't have children, and another version of the quote mentions them. So who knows who he is talking about. It can't be Elon either, because he hasn't changed a thing.
https://www.theverge.com/ai-artificial-intelligence/1029802/trump-60-minutes-ai
Has an app or a site turned your agent away yet? Leave a comment here or reply on Bluesky.