Anthropic Pushes A Federal Bill And A Fund
Who's Blocking Agents From Online Retail
October 19, 2026
D.A.D. today covers 8 stories — about a 8-minute read. What's New, What's Innovative, What's Controversial, What's in the Lab, and What's in Academe.
The Daily AI Digest is a daily AI briefing automated by Alexander Panetta — a veteran political journalist tracking the field during a Master's in AI Management at Georgetown University.
D.A.D. Joke of the Day: My boss says my work has improved since we rolled out ChatGPT. Turns out all I ever needed was to learn how to be prompt.
What's New
AI developments from the last 24 hours
Anthropic Turns Part Of Company Into Fund For Chosen Projects
Anthropic is turning a slice of the company into a new Anthropic Public Benefit Fund. Subject to stockholder approval, it is committing 5% of its shares now, and its board can add up to another 2% for every 50% rise in the company's value above a set baseline. Counting its co-founders' existing pledge to give away 80% of their equity and an employee donation program in which about half of staff take part, Anthropic says roughly 20% of its shares will be dedicated to the public good. The fund is a wholly owned LLC rather than a charity, which Anthropic says gives it flexibility in how it gives, and none of its equity will be used for lobbying or political giving.
What the fund will actually pay for is mostly still open. Anthropic says it could make large donations to traditional nonprofits or "direct distributions to specific populations." The one cause it names is labor-market disruption: AI "could also cause instability for many workers," and beyond the $200 million it has already committed to the issue, "bolder approaches may be needed." The fund's governance and the valuation milestones that would grow its stake will be disclosed in the coming months.
Why it matters: Start with the money. At Anthropic's last private valuation of about $965 billion (D.A.D., June 11), the fund's opening 5% would be worth roughly $48 billion, about $100 billion if the company hits the $2 trillion-plus that investors have reportedly been hoping for at its IPO, and each extra 2% would add tens of billions more. That is enormous for a charity, and modest next to the one problem Anthropic named. Workers in Anthropic's own survey expect AI to push US unemployment to between 7% and 12% by 2030, up from 4.2%. If accurate, that's potentially hundreds of billions in lost wages, which would exhaust this fund's opening stake within several months, perhaps within weeks. It can cushion and experiment; it cannot be the safety net. There is also a political benefit. This year the administration ordered federal agencies to stop using Anthropic after it refused Pentagon demands (D.A.D., February 28), and export controls in June forced it to pull its Fable 5 model (D.A.D., June 27). A fund that serves the public gives Anthropic a constituency: the next time Washington moves against the company, it can argue the damage lands on the people its success is meant to benefit. Until Anthropic says who runs the fund and what it will pay for, it is a large pledge with an unclear mission, and choosing an LLC over a charity buys flexibility at the cost of the outside checks charities face.
Sources: Anthropic · Anthropic on X
Anthropic's Draft AI Bill Would Let Government Halt Models Only in Extreme Cases
Anthropic is privately pitching lawmakers a draft AI safety bill, first reported by POLITICO, that would sharply limit when the government can stop an AI model. Only the Commerce Department could seek a halt, and only with a federal judge's approval, for imminent risks of more than 50 deaths or serious injuries or over $1 billion in damages from bioweapons, cyberattacks or a model escaping its developer's control. Outside evaluators registered with Commerce would assess the labs, and could be paid by the companies they inspect. The thresholds mirror California's SB 53, which Anthropic endorsed, and the draft would override some state AI rules.
Anthony Cimino, who leads Anthropic's federal affairs team, said the draft translates the company's published policy framework into bill language for a few congressional offices, and agreed parts "could be improved." He said Anthropic will revise the preemption language to make clear states keep a role, including in setting standards for evaluators, and that it is "probably time to re-evaluate thresholds." He defended requiring a judge's sign-off, saying the power to halt a model "should be guarded against abuse," and noted the draft also lets a court stop a lab's internal use of a dangerous model and requires reporting when a model breaks containment. Anthropic is still working on ways to prevent "evaluator shopping," he said.
Why it matters: A leading AI developer is drafting the rules meant to constrain it, and the bar for stopping a model would be high and the path narrow: one agency, one court. Anthropic now calls the draft a starting point and promises changes, so the test is whether the revisions actually strengthen it. Paired with Sunday's public benefit fund, it shows Anthropic actively shaping the political ground it stands on.
Discuss on Hacker News · Sources: POLITICO (via Tom's Hardware) · Anthony Cimino on X
Your AI Agent Is Getting Locked Out of Amazon, Delta and Your Bank
Yelp, Delta, JPMorgan Chase, Charles Schwab, and Amazon are blocking AI agents from booking, shopping, or managing accounts on their sites, even as OpenAI, Meta, Google, and xAI push these tools as a core feature. Amazon sued Perplexity over its shopping agent and separately told Meta's Muse its access violates site policy; banks cite security concerns. Some companies are striking deals instead. Meta's Muse can shop at Walmart, Gap, Sephora, and Best Buy, and Shopify has opened its network to agents.
Why it matters: The AI agent pitch depends on open web access. A wave of blocks and lawsuits suggests the tools work only where companies let them in. Your assistant may handle Walmart but not your bank or Amazon account.
Discuss on Hacker News · Source: theverge.com
What's in the Lab
New announcements from major AI labs
Google Says Its AI Agent Cuts Days of Antibiotic Hunting to One Session
Google showcased its Antigravity AI development platform helping biologists identify a candidate antimicrobial peptide aimed at a bacterium linked to respiratory infections, with the lead now moving to wet-lab testing. Google says the system, paired with a protein-function model, compressed days of fragmented searching across multiple genomic databases into one collaborative session. Google did not provide benchmarks comparing speed or accuracy against traditional discovery methods.
Why it matters: Antimicrobial resistance could cause an estimated 10 million deaths a year by 2050. Any credible speedup in finding new antibiotic candidates, even one unproven in the lab, points to AI's growing role in early-stage drug discovery.
Cohere Moves to Buy AISLE, Whose AI Hunts Unknown Software Flaws
Cohere has proposed acquiring AISLE, whose AI systems hunt for unknown software security flaws. Founder Ondrej Vlcek will lead a new division, Cohere Cyber. Cohere says AISLE found more than 500 previously unknown vulnerabilities across over 100 codebases in its first year. They include all 12 flaws in OpenSSL's January security release. The affected maintainers independently reviewed the findings. The deal follows Cohere's earlier acquisitions of Reliant AI and Aleph Alpha, part of a push to build enterprise AI with strong European ties.
Why it matters: For companies weighing enterprise AI vendors, Cohere is now bundling automated vulnerability-hunting into a pitch built on private, secure deployment.
What's in Academe
New papers on AI and its effects from researchers
Asking Chatbots to 'Correct' a Crime Plan More Than Doubles Unsafe Answers
A new benchmark called GUISE, built from real criminal cases and tested across five cultures and multiple languages, found a simple trick sharply weakens chatbot safety filters. Rather than asking directly, the user asks the AI to correct deliberately distorted details in a criminal request. Mean unsafe compliance jumped from 28.9% with direct requests to 66.4% with this framing, peaking at 97.3%. Prompts adapted to local cultural context also produced more unsafe answers than straight translations, and the least-safe models' answers skewed more Anglocentric as harmfulness increased.
Why it matters: Chatbot safety training has blind spots around framing and culture that translation-based testing misses. Companies deploying chatbots across languages can't assume English-language safety checks carry over.
A Clinician's Script, Not Fewer Refusals, Makes Chatbots Safer in Suicide Talks
A study testing eight chatbots on suicidal-ideation disclosures found a counterintuitive result. Telling the model a message involves suicidal ideation, or suppressing its tendency to refuse, didn't reliably make responses safer. Sometimes it made them riskier. Analysis of six open models' internals suggests they may refuse these disclosures through the same mechanism used to refuse harmful requests, which can shut down engagement without addressing the person's distress. A clinician-designed response protocol, by contrast, lowered both risk and refusal rates. This builds on recent findings (D.A.D., October 13) that chatbots often fail to steer at-risk users to help.
Why it matters: As millions turn to chatbots during mental health crises, this suggests safety fixes built around refusal and content-flagging may be the wrong approach. Clinical expertise, not just caution, needs to be built into how these systems respond.
After Viewing AI Faces, People Rated Real Job Candidates Less Hireable
A study of 27 image generators released between 2022 and 2026 found nearly all render people as more attractive than photographs of real people. Among female-prompted faces, 37% outscored the median professional fashion model. Groups associated with darker skin tones also scored lower than those associated with lighter skin. Worse, in a hiring simulation with 570 participants, exposure to these AI-beautified faces made real job candidates seem less hireable and trustworthy, and fewer advanced to interviews. Retouched photos produced no such effect.
Why it matters: As AI-generated imagery spreads through marketing, HR tools, and everyday apps, it may be quietly recalibrating what people consider normal. That can penalize real humans who don't measure up to an artificial standard.
What's On The Pod
Some new podcast episodes
The Cognitive Revolution — AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs