Daily AI brief · October 7, 2026
Meta and Walmart Back a Standard for AI Agents
Sierra and Meta published an open standard for how personal AI agents sign in to businesses, ChatGPT was caught forging real cartoonists' signatures, and an AI job interviewer went on general sale.
Meta, Sierra, Walmart and Stripe publish a standard for personal AI agents (The Next Web, PYMNTS, CMSWire, Sierra) Sierra, the customer-service agent company run by Bret Taylor and Clay Bavor, announced the Personal Agent Protocol yesterday with Meta, and named Shopify, Stripe, Walmart, Genesys, Instinct and Rocket as partners. It describes how a customer's own AI assistant finds out what a business offers, browses as a guest to check stock or a returns policy, and then, when the customer signs in, gets either read-only or write access, using the same sign-in technology most websites already rely on. A first version of the specification is due later this month, and payments are left for a later stage. For a small business, this is the clearest sign yet that some of your "customers" will soon be software acting for a person, and the businesses that have already decided what such an agent may look up, change or order will handle it calmly instead of scrambling.
ChatGPT is forging real cartoonists' signatures (Nieman Lab) Nieman Lab found more than 15 New Yorker cartoonists whose signatures or pen names ChatGPT has stamped onto AI-generated cartoons when users ask for "a New Yorker-style cartoon," including Brendan Loper's "BLOPER" on a cartoon about Dolly Parton that spread on social media. OpenAI's response so far is a warning message about similarity to third-party content. If you run marketing, read every AI-made image for names, logos and signatures before it goes on your site or social feed, because the tool will not always tell you it borrowed someone else's work, and the person whose name appears on it will not care that a chatbot did it.
HackerRank puts its AI job interviewer on general sale (TechCrunch) HackerRank made Chakra, its AI interviewer, generally available after a six-month beta that ran about 500,000 interviews, with Snowflake, Snorkel and Capgemini among the early customers. The company says it scores how a candidate reasons and can fold a phone screen, a take-home task and a panel into one sitting. Whoever does the hiring in your business will be pitched tools like this soon, and the right first question is not how fast it screens but how you would explain to a rejected candidate, or a regulator, why the tool scored them the way it did.
Microsoft's first big Windows event in over two years is today (Thurrott, Windows Latest, TechSpot) Microsoft holds a Windows and Surface event in San Francisco at 1 PM Eastern today, with Satya Nadella and Nvidia's Jensen Huang expected, and the theme is AI that runs on the PC itself rather than in the cloud. A new Surface Laptop Ultra built on Nvidia hardware is widely expected. The practical part: if you are about to replace a batch of office laptops, wait until you've seen what Microsoft says today about which machines get its new on-device features, so you don't buy a fleet that misses them by a few months.
Write the agent rule before the agent shows up. Have your AI champion spend twenty minutes on one question: if a customer's AI assistant signed in to your website or booking system, what could it look up, and what could it change or order? Write the answer down in plain words. You don't need to build anything yet, but the businesses that have an answer will be the ones that say yes to this traffic safely.
Add one line to your staff AI rules. "Before any AI-made image or text is published, someone checks it for names, signatures, logos and claims we can't back up." It takes a minute to add and saves you the awkward email from an artist or a competitor.
Be wary of AI interviewers you can't question. If a hiring vendor offers AI screening, ask to see the scoring for a sample candidate and how a person can override it. If they can't show you that, skip it for now, because you, not the vendor, will be the one answering for who got turned away.
Where it's going: The big companies are now writing the rules for how AI agents deal with businesses, and the small businesses that decide their own limits early will get to follow those rules instead of being caught by them.