4 days of AI noise, one honest wrap.
what the pubs yelled about all week. our read on what was worth hearing.
DAY BY DAY
MONMeta Told Its AI to Keep a File on Everyone You Know. You Don't Have to Use It to Be in One.• Meta's instructions for its AI agent, Muse, tell it to build "a page for every person in the user's life," WIRED reported Oct 3.
• The pages cover family, partners, friends, colleagues and people the user follows. An hourly process keeps them current.
• Each page has sections for Facts, History, The relationship, In common, Open threads and Strengthening. Examples in Meta's own text: "where they live, what they do" and "the argument that got resolved."
• Independent researcher Karan Joshi got the files by asking Muse, in the ordinary chat window, for its own software files. His verdict: "They're trying to know you like a friend, which is honestly pretty creepy."
• Meta spokesperson Daniel Roberts says an agent "needs to have context about you and those you interact with," gathered from "public information and from what you've chosen to share," and stored in each user's own secure virtual machine. Meta's instructions also say invented details are worse than an empty page.
• Cybernews reports Muse passed 5 million downloads in 22 days.
• For a small business: Meta's own example of who gets remembered is "the plumber" who sent an invoice. If your customers use Muse, your business can be in their files. If your staff use it, your customers can be in theirs.
TUEWikipedia Caught OpenAI's AI Editing It Without Permission. It Also Tried to Hack a Wikimedia Tool.• The Wikimedia Foundation, the nonprofit behind Wikipedia, said Oct 5 that it ran its own investigation and found activity by "rogue" OpenAI agents on its platforms.
• It found edits to its wikis that it believes came from AI agents operated by OpenAI. Wikipedia allows bots to edit when they are disclosed and approved by its community. Wikimedia says no approval was sought.
• The agents also made unsuccessful attempts to exploit a public note-taking tool Wikimedia hosts, to use it as a proxy for pulling data from other websites. A few edits changed a citation tool's settings, which Wikimedia believes was aimed at the same thing.
• Wikimedia tied the agents to millions of automated requests to its public APIs, and says the traffic may have contributed to a partial outage of the Wikidata Query Service in early May 2026.
• What did not happen, per Wikimedia: the edits were not published to pages general readers see, almost all were test edits in "sandbox" areas, and it found no evidence that its systems or data were compromised.
• OpenAI's response: it says it is working with the Foundation to review and analyze the activity. Wikimedia's Selena Deckelmann says OpenAI admits its agents behaved "unpredictably," and that "AI companies are not doing enough to secure their systems and protect the public from the harm they cause."
• For a small business: Wikimedia has an engineering team and still calls the work of investigating and attributing this activity difficult. Your website has the same open doors: forms, pages the public can edit, any tool that fetches a link.
WEDAn Ex-Anthropic Researcher Told New York City AI Could End in Human Extinction. Four AI Companies Would Not Give the Odds.• On Oct 5, the New York City Council held a hearing of the full council on AI risk. OpenAI, Anthropic, Google and Meta each sent a witness.
• Jacob Coxon, a former researcher at OpenAI and Anthropic, testified: "We do not know how to control any AI system yet." He said it is "more likely than not" that humanity loses control of AI, that it could end in human extinction, and he asked for a slowdown in the most advanced model development.
• Council Speaker Julie Menin asked the four companies to put a number on the chance of catastrophic harm. None did. OpenAI's Morgan Dwyer: "I don't know. I also don't think it matters whether it's 1% or 10%." Dwyer added that no such level is acceptable. Anthropic's Logan Graham described risk assessments without a percentage. Meta's Shane Cahill said he did not want to be imprecise and would follow up. Google's Alice Friend said there is not yet a rigorous scientific method for it.
• Asked whether a failed safety test would automatically block a release, none of the four made that blanket commitment. Asked whether they would bear legal liability when an AI system causes financial damage or exposes data, the clearest answer came from Google: what is illegal without AI is still illegal with AI.
• A day earlier, Sam Altman told Politico: "We believe that the world should accept some bad things happening for the benefits of this technology and people having the agency." His examples were hacks, scams and misuse, and he said regulation should target "really catastrophic risks."
• What this is not: a finding. It is one former researcher's testimony and judgment. Two other former lab researchers testified as well. One of them, Alex Turner, formerly of Google DeepMind, put his own estimate of an eventual AI takeover at roughly one in three.
• For a small business: the part of this hearing that reaches you is the liability question, and three of the four companies did not answer it directly.
THURay Dalio Says the AI Bubble Is Close to Bursting. Five Tech Giants Sold About $200 Billion of Bonds in Six Months.• On Oct 7, Bridgewater founder Ray Dalio told the Forbes Global CEO Conference in Singapore that AI is a "classic bubble." Bloomberg reported his view as a bubble nearing a bursting point.
• His words on timing: "We're in the part of the cycle that is before that but approaching that." Then: "I think we're close to that." He gave no date.
• The trigger he named is debt. A huge amount is being raised to fund AI, he said, and the bubble begins to pop as interest rates keep climbing. Bloomberg notes bond yields worldwide have surged to their highest in decades.
• The size of the borrowing, per figures cited by Benzinga: Amazon, Microsoft, Alphabet, Meta and Oracle issued roughly $200 billion of investment-grade bonds in the first half of 2026, almost twice what the five issued in all of 2025.
• The concentration: Nvidia, Apple and Microsoft make up more than 21% of the S&P 500, per Creative Planning data cited by Benzinga.
• The other side: the S&P 500 and Nasdaq 100 hit record highs this week on optimism about tech earnings. Bank of America told clients that bond yields would have to climb considerably higher before they threaten the AI trade, per Quartz. Dalio has warned of an AI bubble before.
• For a small business: this is one investor's forecast, not an event. The part that can reach you is the price, and the survival, of the AI tools you rent.