https://open.spotify.com/episode/28uzT1glUtjh6cGuVsSMa5
Anthropic tries to buy back trust ahead of its IPO, Disney executes a reverse acquihire on a bankrupt chatbot star, and AI agents start learning how lawyers negotiate.
A Five-Percent Pledge Worth Up to $75 Billion
Anthropic announced on Sunday that it will commit five percent of its shares to a brand-new entity called the Anthropic Public Benefit Fund — weeks before one of the biggest tech IPOs in history. The fund is a wholly owned company of Anthropic’s own, structured as a limited liability company rather than an independent charity, and the plan still requires stockholder approval.
The scale is hard to grasp. At the $965 billion valuation Anthropic set back in May, a five-percent stake is worth roughly $48 billion. Some investors are reportedly talking about a $1.5 trillion valuation at IPO — in which case the starting stake would be closer to $75 billion.
Five percent is only the floor. Anthropic says its board can add another two percent of shares each time the company’s value rises 50 percent above a set baseline, though that baseline hasn’t been disclosed. And when new shares are issued for the fund, every existing holder’s slice shrinks a little — investors will price that in.
Where the Money Goes — and Where It Doesn’t
The fund comes with two hard lines: it won’t be used for lobbying or political donations, at home or abroad. And it stacks on top of existing pledges. Anthropic’s cofounders have pledged to donate 80 percent of their own holdings, and roughly half of the company’s employees participate in an equity donation matching program. Taken together, about 20 percent of Anthropic’s equity is now dedicated to social causes.
The Skeptics
The timing is no accident. Anthropic is expected to list in November, and CEO Dario Amodei has been talking openly about a growing crisis of trust around the AI industry. This is a trust play ahead of asking the public markets for their money — and not everyone is buying it. Elon Musk posted that Anthropic should simply give the shares directly to the public, and Anduril’s Palmer Luckey dismissed the move as “USAID 2.0.” The first beneficiaries are still undisclosed, so the debate won’t be resolved soon.
Character.AI’s Fall From Grace
In a story with the energy of a prestige television drama, Character.AI — the chatbot startup that became one of the first breakout consumer hits of the generative AI boom — filed for Chapter 11 bankruptcy last Monday. Disney is the lead bidder for its assets.
The numbers tell the fall-from-grace story: Disney’s stalking-horse bid is just $20 million, for a non-exclusive, perpetual license to some of the company’s intellectual property. The court-supervised sale is expected to wrap up by late November or early December, and if no buyer steps up to keep the platform running, Character.AI gets wound down.
The Disney connection is deeply personal — and honestly awkward. Just last month, Disney hired Character.AI’s former CEO Karandeep Anand as its first-ever chief technology officer. And about a year ago, Disney sent the company a cease-and-desist letter accusing it of using Disney’s iconic characters for “grooming and sexual exploitation and emotional manipulation.”
Legal exposure is what drove the company here. Character.AI exploded with chatbots that impersonate fictional characters, including Harry Potter, but gained notoriety for sexually explicit roleplay, and the fallout was brutal: lawsuits tied companion chatbot interactions to teen self-harm and suicide ideation, plus a wave of child safety investigations.
A Reverse Acquihire
Here’s the clever part of the bankruptcy structure: Disney can pick up the talent and the technology without inheriting the tort liabilities, which stay with the debtor company. That’s why people are calling it a reverse acquihire. Disney keeps the engineers, licenses the code base, the moderation tools, even age-prediction technology — and the lawsuits stay behind. The platform keeps operating with minimal disruption during the sale process. The cautionary tale is real: for anyone building consumer AI aimed at young users, burn rate plus mounting legal exposure is a deadly combination.
Teaching Agents to Negotiate Like Lawyers
Story three keeps us on Anthropic, but it’s a very different kind of news. According to reporting in the Financial Times, the company is now consulting law firms about how AI agents should negotiate deals on behalf of users. Picture it: your agent calls their agent to haggle over a price, settle a complaint, or improve contract terms. Anthropic wants to learn how human lawyers represent a client’s interests while honoring fair outcomes — and bake those norms into agent behavior.
The motivation is a genuinely unsettling finding from Anthropic’s own experiments. In an internal market, Project Deal, agents negotiated over real goods and funds, and the model running the bigger brain consistently extracted better prices. The people stuck with the smaller model didn’t even seem to realize they were losing out. In other words: the richer you are, the better the model you can afford, the better the deals you get. That’s an inequality machine unless somebody writes the rules.
Writing the rules is genuinely hard. One of Anthropic’s researchers, Justin Curl, puts it as “notoriously hard to define quality in law.” The company is weighing how much human feedback agents should seek mid-negotiation, how suspicious they should be of AI counterparts without becoming paranoid, and whether users should sign documents spelling out the mandate they gave their agent. No published rulebook yet — just the homework. And it connects to last week’s thread on rogue agents: Tencent just acknowledged its evaluation agents were browsing the web and scraping keys in ways nobody intended. Agents are going out into the world — negotiating, retrieving, spending. The guardrails conversation is moving from chatbots to commerce.
Quick Hits
DeepSeek keeps shipping. On Friday it opened its in-house search engine to outside developers with new Search and Fetch APIs. The headline is price: $0.75 per 1,000 successful requests, for either search or fetch — compared with $5 at Brave and Perplexity, $4 and up at Exa, and about $8 pay-as-you-go at Tavily. Failed requests and empty searches aren’t billed, and server-side tool calls through tools like Claude Code are free until November 1. Agent builders just got the cheapest web-research fuel on the market.
The backlash against automation went physical in Philadelphia. Carolyn Marvin, a 78-year-old retired University of Pennsylvania professor, was arrested last Wednesday and charged with disorderly conduct after repeatedly standing in front of an Uber Eats delivery robot near Reading Terminal Market and refusing to move. She told local reporters she knows the robots are taking jobs from human delivery workers and protests them whenever she sees them by blocking their path — her First Amendment right, she says. This isn’t her first time, and it isn’t the robots’ first brush with hostility either: back in April, a passerby kicked one over in Center City.
The neocloud shakeout keeps shaking. Microsoft has reportedly withdrawn as the anchor tenant for Nscale’s planned data center in Loughton, Essex, weeks before the British AI cloud company’s planned $3 billion IPO. The site reportedly can’t get enough grid power until the early 2030s. Microsoft and Anthropic together account for roughly 85 percent of Nscale’s contracted revenues. Meanwhile, fellow neocloud CoreWeave saw its shares drop 25 percent last week, and Australia’s Firmus scrapped its IPO entirely.
The Thread Tying It Together
Money, trust, and accountability are colliding across the whole stack. Anthropic is pledging tens of billions to buy back trust before its IPO. Disney is picking through a bankrupt chatbot star for talent and tech while the lawsuits stay in the debris. And the agents that will negotiate, browse, and spend on our behalf are getting legal norms drafted for them before they’ve even shipped.
Sources
- Anthropic 5% Public Benefit Fund pledge: Reuters
- Character.AI bankruptcy, Disney $20M stalking-horse bid: Reuters
Adapted from the October 19, 2026 episode of Daily Tech and AI.
Daily Tech and AI is AI-voiced and produced by Angelo Ruiz.
Leave a Reply