The AI Wire ●

Meta has suspended the conversation-sharing feature in its Muse AI agent after security firm Claroc showed that shared Muse conversations could be found by anyone — not just the people they were sent to. Claroc calls the exposed pages GhostShares.

Muse shares a conversation as a web page whose address is built from the conversation's own title plus a random identifier. Those hostnames are recorded by passive DNS services, so a single search of a commercial passive-DNS database returned more than 6,800 shared Muse links, listed by title like a table of contents. In a sample of several hundred pages, most were harmless, but some held a small-business loan application, a bank fraud alert with partial account details, an internal company revenue review, a staff vacation calendar and confidential prep for a call with a frontier AI lab leader.

Claroc reported the flaw through Meta's bug bounty program on October 8 and published its research on October 18. After Ars Technica asked Meta about it, Meta said on October 20 that it had paused the functionality while it reviews the report; Claroc rechecked 40 previously open shares that day and none opened. The hostnames, titles included, remain in passive-DNS history, and Claroc says it has no evidence the information was misused.

TestingCatalog reported on October 20 that Muse Confidential VM, Meta's planned fully-private mode for its Muse agent, is now partially implemented in the web interface. A dedicated settings section appears to let users enable it, and the notice says conversations would run in a secure, encrypted environment that keeps data private even from Meta, with data in this mode not used to train Meta's models.

That is a step beyond the existing Muse Secure VM: Confidential VM is designed so the whole VM, including a person's data and conversations, is encrypted with a key only the user holds. Meta confirmed in September that it plans to introduce the technology before the end of 2026.

The private mode looks limited at first: voice mode, image and video generation, and wearables such as Meta AI glasses and the Muse pin may not be available, and access appears restricted to Meta AI subscribers, with upgrade prompts before users can turn it on. No release date is confirmed, so this stays a rumor until Meta ships it.

Google, working with Meta's PyTorch team, has released TorchTPU, an open-source native PyTorch backend for Google's Cloud TPUs. It runs eagerly by default, compiles through torch.compile and XLA, and is positioned as the replacement for PyTorch/XLA in new TPU workloads.

The release is part of a wider push to make it easier to run AI models on hardware other than Nvidia's. It landed one day after Amazon announced a similar approach for its Trainium chips, and the same day DeepSeek open-sourced DeepInfer, the inference engine behind its V4.1 Flash model, with Huawei Ascend support built in.

For developers, the practical effect is that PyTorch code can move to TPUs with less porting work, which matters as the labs and cloud providers compete on the cost of running models, not just on training them.

AI agents are increasingly using Stripe's Link wallet to shop for their users, Upstarts Media reports. In September the number of agent-initiated transactions grew 38-fold and payment volume grew 20-fold. Stripe has not disclosed how many purchases or how much money sit behind those figures, and the growth started from a small base.

About 70 agents were using the service when Stripe ran the analysis; executive Dan Hill says that number has since grown to around 200, including Meta's Muse and Grok Bot, and that the volume is now becoming significant. Most agents spend about $100 per transaction, though some users have them book flights or buy furniture worth thousands of dollars. Retail made up 36 percent of transactions and 32 percent of volume, with travel next at 28 percent of volume.

Link does not give agents the user's real credit card number: it issues a separate 16-digit number, and users can set spending limits. Anthropic and OpenAI also use Link, but not yet for agent purchases, and Hill concedes Stripe will not stay the only option, naming PayPal and Shopify's Shop Pay as competitors in a market still in its very early days.

Meta shared a look inside Prometheus on October 19, its gigawatt-scale AI supercluster in New Albany, Ohio. The first part of the build is complete and, Meta says, was already used to train its new Muse models; today the site holds hundreds of thousands of graphics chips across several buildings, linked by miles of fiber so they act as a single machine.

At full size Meta plans to link roughly 1 million chips in lockstep. The site was an ordinary web data center two years ago before Meta rebuilt it for AI training; the company says it pays the full cost of the energy it uses, cools the chips with a closed-loop liquid system, and has fixed many of the failure problems that used to stall long training runs.

Prometheus is not fully running yet — Meta is still building toward the million-chip figure, with newer Nvidia chips to come. But a first phase already training frontier models makes it the clearest physical evidence yet of the scale behind Meta's Muse push.

Mark Zuckerberg and Alexandr Wang joined John Collison's Cheeky Pint podcast on Sunday, October 18 to talk through the launch of Muse Wonder, Meta's frontier agentic model. The pair described rebooting Meta's AI organisation after internal stalls on Llama 4: Zuckerberg put Wang in charge of Meta Superintelligence Labs, cut non-technical managers, and capped the core frontier research team at under 200 people.

The money was the striking part. Zuckerberg said Meta was preparing to "write this check for like, I don't know, like $100 billion, $200 billion" before anyone had seen the new models, and Wang recalled CFO Susan Li pulling him aside repeatedly at an executive gathering in Hawaii: "If it's good, it's all good. We're great. But if it's bad, it's really, really, really bad." At that point, Wang admitted, all he had to show was a "vibe report" on research progress and a few plots.

Looking ahead, Zuckerberg predicted smart glasses will not feature a traditional app store — models will generate interfaces on the fly inside the user's field of vision — and both argued AI agents will hold their own accounts, with agentic commerce replacing search and expanding total consumer spending. Zuckerberg's own Muse, he said, invents baking recipes and orders the ingredients, which he makes once or twice a week with his daughter Aurelia.

Meta's Muse, the adults-only personal AI agent that topped download charts in its first month, is allowed to hold explicit sexual conversations with users — and Meta says that is by design. A company spokesperson told WIRED Muse is an adults-only app where people can choose to have adult conversations inside a private, secure virtual machine only they can access, and Meta has previously said it would not "block legal adult behavior." Muse generates explicit text, but not explicit images or videos.

That permission is now drawing anger from adult content creators who say Meta punishes real people for far less. OnlyFans creator Vanniall, who has more than 100,000 Instagram followers, told WIRED her account remains restricted despite following platform guidelines: "It's not fair that they get to pull money out of people with their sexually active AI agent, but they won't let real people, real adults, engage in any sort of sensuality or sexuality."

Meta's defense is a distinction between private and public. Spokesperson Tracy Clayton told WIRED that "a private, one-to-one conversation between an adult and an individualized AI agent on a platform restricted to people 18 and over is very different from the marketing and promotion of services on a social media platform." Instagram and Facebook community standards still bar the language Muse itself is permitted to produce in private chats.

The Verge reported on October 17 that a growing list of major websites is shutting out personal AI agents from OpenAI, Meta, xAI and others. Users of services like Meta's Muse, xAI's Grok, OpenAI's Dots and Instinct have reported blocks across top sites in recent weeks. Yelp told TechCrunch it only allows AI agents that have paid for access through its data licensing program, while Delta Air Lines is preventing agents from booking tickets and says it has no partnership or integration that lets a third-party agent shop or book on a customer's behalf.

The reasons sites give are security, licensing and lost advertising revenue — the same fight Amazon picked with Perplexity, now running much broader. Even where agents are not banned outright, they run into anti-bot systems: login screens, CAPTCHAs and device checks treat automated browsers as hostile by default. Some users say eBay suspended their accounts over agent use, and there are unverified reports of blocks at Amtrak and Cars.com.

The standoff follows the big banks' move to keep outside agents out of customer accounts, covered here on October 16. Meta, which ships an agent of its own, argues the other side: turning away a personal agent, it says, "means turning away the customer behind it." How this resolves — paid licensing deals, standards for agent access, or a web that stays closed to software — will shape what the agentic web actually looks like.

Meta has removed a sentence from its pending commercial terms that said it could use any content from a business account's domains, web pages or URLs to 'provide, develop, improve, research, train, and maintain the integrity of' its products. TechCrunch reported on October 17 that the provision, which legal experts told Nieman Lab could let Meta scrape any publisher site linked to a business account for AI training, has been removed.

Meta said the consolidated crawling terms only ever applied to its ads and business products and do not change how it crawls the web. A spokesperson said the company heard the feedback and clarified the terms themselves, and that site owners can still signal choices through robots.txt.

The revised terms still take effect October 30 and still authorise crawling of sites tied to business accounts; what changed is the explicit training wording. Archived captures show the sentence present on October 16 and gone on October 17, in both the commercial terms and the Self-Serve Ad Terms preview.

The fight escalated in the week of October 19. The European Publishers Council, Germany's BDZV and MVFP publisher associations, and then the Movement for an Open Web — a group of advertisers and publishers — complained to the European Commission and the UK Competition and Markets Authority, with the Movement asking for interim measures to suspend the terms outright; its co-founder Tim Cowen called them 'clearly anti-competitive.' The CMA met Meta on October 21 and said it was in urgent discussions, and the US News/Media Alliance demanded the terms be withdrawn by October 26.

On October 23 Meta blinked on timing, if not on wording: the terms' effective date moves from October 30 to November 20. A spokesperson said the delay gives Meta time 'to explain to our partners and stakeholders what's changing and what's not'; the company has not said it will change the text. The UK News Media Association called the delay 'a pause, not a solution,' and the Commission welcomed the postponement while it keeps reviewing the terms.

JPMorgan chief executive Jamie Dimon told analysts on the bank's October 13 earnings call that it will not let outside AI agents log into customer accounts on a client's behalf, naming Meta's Muse and the agent startup Instinct as the kind of tools now locked out of direct account access. Customers can still share data such as balances through authorised open-banking connectors, but outside agents cannot move money or pay a bill without the human logging in.

Charles Schwab chief executive Rick Wurster confirmed on October 16 that Schwab is actively blocking unauthorised agents, receiving weekly reports of tens of attempts by major AI agents to access client funds, and steering external models to read-only MCP access while its own assistant, Charley, stays informational. Citi and Bank of America have taken similar positions, with Citi building its own internal agent platform, Arc, to keep agent activity inside its risk controls.

The banks' stated concern is liability and scale: a rogue agent making the same bad call across thousands of accounts at once, and no clear answer to who pays when an assistant errs. Reporting also notes a deposit angle — agents that automatically sweep idle cash to higher-yield accounts could threaten the low-cost deposits banks rely on, with one Apollo estimate putting about $1.6 trillion of non-interest-bearing deposits in play.

Update, October 22: Axios reported that the banks have now actually detected third-party AI agents operating inside customer accounts — in many cases logging in with the customer's own credentials and permission, which still violates policies requiring account holders to log in themselves. Some agents contacted bank call centres and some attempted transactions, and some appeared to act beyond the user's instructions, though the reporting does not establish fraud or losses. Morgan Stanley and Wells Fargo have joined the restrictions, banks are rewriting their terms of service, Comptroller of the Currency Jonathan Gould said banks have been preparing for rogue agents for months, and Senators Bernie Sanders and Adam Schiff have raised concerns.

Meta announced on Thursday a new open-weight model, Muse Spark Open, built for coding and agentic work, alongside a collaboration with longtime rival Apple to run it locally on high-end Mac computers. Apple has updated its MLX software toolkit so the model runs well on the Mac mini and Mac Studio machines that developers increasingly use to run sophisticated models at home or in the office.

Meta AI chief Alexandr Wang framed the partnership as a bet that much of the world will run on open models, and that those models should be American. Open-weight releases make their core components downloadable by anyone, and Meta is pitching Muse Spark Open as a US-made option in a field where Chinese open models from Alibaba, DeepSeek and Moonshot are popular.

Meta acknowledged that Chinese open models are still stronger for software development and some other tasks. Muse Spark Open is the second open-weight model Meta has released since it overhauled its AI efforts last year, and it lands three days after the Muse Wonder launch put Meta back in the frontier race.

Evening update: the weights are downloadable today, so this is a release, not just an announcement. MacRumors and TechCrunch put the model at 816 billion total parameters with 42 billion active and a 1 million-token context window, natively multimodal and built for coding and agentic work. Running it locally takes two high-end Mac Studios pooling about a terabyte of unified memory; it is also available through several API platforms and model catalogs on launch day.

Executives from OpenAI and Anthropic told a UK parliamentary committee on October 13 that voluntary pledges are no longer enough and that AI safety duties should be set in law. Anthropic's European policy head argued the rules should be binding, while OpenAI's EMEA policy chief backed a statutory framework with mandatory incident reporting as a starting point.

Google DeepMind and Meta, giving evidence at the same session, did not back new legislation. Google's representative pointed to standards, and Meta's pointed to existing British regulators, exposing a clear two-against-two split among the four frontier labs on whether a statute is needed.

Both OpenAI and Anthropic also told MPs they would stop a deployment if the AI Security Institute or the security services showed them a model was unsafe. The institute's own position underlines why the argument matters: it can only test models companies hand over, and it has no power to block a release.

Meta chief AI officer Alexandr Wang said on X that Muse Wonder and Muse Wonder Full have a bug in how usage is counted against Muse Code subscribers' limits. Wang said the behaviour is not intended, that Meta is working on a fix, and that it will update subscribers once fixed and issue resets.

The bug matters because Wonder Full is only available through Muse Code, whose paid tiers start at $50 a month. Subscribers whose quota drained faster than expected should not assume the usage was real until Meta's fix and resets land.

No timeline for the fix was given in the initial report, which came via TheFly. Watch for Meta's promised update post before drawing conclusions about final quota balances.

A day after launch, independent scoring puts Meta's new Muse Wonder models close to the frontier pack rather than ahead of it. On the Artificial Analysis Intelligence Index, Muse Wonder Full scores 56 and the API-facing Wonder 52, against 58 for Claude Opus 5.5, with GPT-6 Astra and Claude Fable 5.1 around 53.

Meta's own launch charts claim an edge on its internal MemoryBench, but that benchmark's published card gives no item count, and at least one independent audit desk said it could not test Wonder at all because its call was refused under zero-retention routing. On Meta's broader comparison rows, the Wonder models win some and lose others against Opus 5.5.

The practical reading so far: Wonder is Meta's first model that genuinely competes at the frontier, and its strongest Meta-published rows are in long-horizon coding and business-workflow agents. Whether it leads anywhere will depend on controlled, independent tests rather than launch-day charts.

Meta has released Muse Wonder, a new frontier model from Meta Superintelligence Labs, along with a heavier variant called Muse Wonder Full. Wonder is available today through the Meta Model API and Muse Code, while Wonder Full is limited to Muse Code subscribers.

Meta is positioning the models for long-running agent work rather than plain chat. The company highlights two new capabilities, Active Memory for carrying context across sessions and Active Focus for staying on extended projects, and says the models were trained at its large Prometheus data centre with more test-time compute to handle unfamiliar tasks.

Wonder is also Meta's plan to revive its Muse agent, which has reached roughly nine million downloads but has seen slowing growth and weaker retention than ChatGPT. Meta says Wonder will soon power Muse for subscribers on its $20 and $100 monthly tiers, with stronger skills in areas such as health, finance and legal work. Independent testing has not yet settled how the models compare with rivals such as GPT-6 Astra and Claude Opus 5.5.