❝

It's Friday, November 20th: Welcome to The Stress Test 🔍

Anthropic just went public promising its safety mission comes first. This week's Stress Test looks at who actually has the power to enforce that, and what could change it.

🔍THE STRESS TEST

What Changes When Anthropic Answers to Wall Street

One safety story a week, pressure-tested for what's actually happening underneath the surface.

Yesterday, Anthropic started trading on the Nasdaq. Shares priced at $83 opened at $111.50 and closed about 17% higher, in what was, counting a private placement, the largest IPO on record.

An IPO is the first time a company sells shares to the public. After that, anyone can buy in, and the stock price becomes a public scoreboard that reacts to every decision. When a company goes public, it has to hit its numbers every quarter, and the stuff that doesn't make money right away is usually the first thing to go. For example, at an AI company, that could mean rushing a model out before it's been fully tested. But Anthropic says it set itself up so Wall Street investors don't get the final say on that.

The headline version is that Anthropic locked Wall Street out of its safety decisions. Let's stress-test the headline.

Start with what's protecting the mission. Anthropic is a public benefit corporation, so its directors are legally required to weigh its mission alongside profits. On top of that, a trust of five people with no financial stake holds a special share that lets it elect the directors who make up a majority of the board, per the IPO filing. The filing even warns investors that some decisions "may not maximize stockholder value." When Google went public in 2004, it told investors its motto was "Don't be evil," but nothing legally held the company to it. Anthropic's commitment is written into how the company is set up, which is a real step up from a slogan and gives some reason for hope.

Then look at who else holds power. The founders control a separate share carrying 50.1% of the vote. The trust can't make everyday decisions, UC Berkeley law professor Stavros Gadinis told the AP, so its main tool is replacing directors. And its power can be changed or ended with approval from the board and trustees, or from large shareholder majorities. Anthropic says that's deliberate, to "preserve long-term flexibility if circumstances change." It takes big approvals, not one unhappy investor, so the trust is durable, not permanent.

Now ask when any of it gets used. I used to audit social media platforms for compliance with Europe's Digital Services Act, and the lesson that stuck was that being compliant didn't mean being safe. Governance is the same. A structure tells you who can say no, not whether they will. As Sarah Kreps, who directs Cornell's Tech Policy Institute, told the AP, the real test comes "if and when the question of public benefit becomes costly."

So the protections are real, and they go further than Google's 2004 promise. But we won't know how well they hold up until the first time safety costs Anthropic money.

The thing to watch is how Anthropic talks about safety tradeoffs on its first earnings calls, and whether any change to the trust's powers ever shows up in a filing. Now that it's public, we'd all see it.

Each week, we highlight AIC chapters doing groundbreaking work with their members around the world. Tag us on socials to be featured!

🇧🇩 Bangladesh | Hack for Humanity Finale Crowns an AI Recycling Platform

Image from The AI Collective Bangladesh Chapter

The AI Collective Bangladesh Chapter wrapped up Hack for Humanity: Bangladesh, a national hackathon on protecting the environment with AI. The competition began with 179 registered teams, narrowed to 30 working prototypes, and sent 12 finalists to the Grand Finale at American International University-Bangladesh.

Notun Alo, a reward-based recycling platform from University of Liberal Arts Bangladesh, took the championship, ahead of NeuroFlight Lab’s JolRekha: The Water Line and Social Catalyst’s Reroute Agro, which routes perishable vegetables to market. Thank you to the sponsors, mentors and volunteers, and to the Bangladesh team for leading the chapter.

🇦🇺 Canberra | New Chapter Tests Whether AI Can Read Your Website

Image from Eric Nguyen

AI Collective Canberra is one of our newest chapters, and organizer Eric Nguyen showed what its members have been building. At a panel on How to Control the Synthetic Narrative, hosted by Medianet and IABC Canberra, he shared two experiments: a microsite prototype members built in three days that both ChatGPT and Google AI Overviews cited, and simulated digital personas that tested website journeys and surfaced usability gaps.

His bottom line: communication is now about how AI interprets and represents what you publish. The chapter made its first mark earlier this year with a Hack for Humanity hackathon for local volunteer groups.

🎙️ This Week on The AI Collective Podcast

Week five of the show, hosted by AJ Green and Catherine McMillan, moved from the future of entry-level work to the creator economy to how AI should be taught. On Monday, Mickey McManus of Boston Consulting Group joined, alongside Arsh Shah Dilbagi. Mickey raised a question that stuck with the team: as AI absorbs entry-level work, what happens to the apprenticeship model that turns beginners into experts?

On Tuesday, Allan Ko, founder of Influenxio, argued that AI won’t replace real creators because trust and authenticity can’t be automated. Instead, he said, it should handle the work behind the scenes, like brand deals and campaign execution, so creators can focus on creating. Wednesday’s show ran as a Collective Intelligence Live edition. On Thursday, Sinan Ozdemir, who leads AI developer education at Fireworks AI, joined AJ Green. Sinan has worked with language models for more than a decade, founded and sold a Y Combinator-backed AI company, and written around ten books on machine learning and LLMs. His approach is to teach the parts of AI that don’t change from release to release, so people can judge new models for themselves, and he was welcomed into our Hall of Fame.

Today at 9 AM PT, Episode 25 goes live. The show is live every weekday at 9 AM PT on LinkedIn, YouTube, X and Instagram, and every episode is on our YouTube channel.

🫵 Want your message in front of 500,000 AI builders?

Our partners and sponsors get exclusive placements across the newsletter and access to AIC's in-person network: demo nights, dinners, hackathons, and forums across 180+ chapters.

For all inquiries, send us a note at [email protected].

The AI Collective is built by volunteers across 180+ chapters in 40 countries.

Thank you to the thousands of volunteers around the world who make this work possible. We truly could not do this without you.

🧑‍💻 About the Editors

About Joy Dong

Joy is a news editor, writer, and entrepreneur at the intersection of AI and blockchain. In her newsletter TEA, she demystifies complex systems, and her mission is to make emerging tech accessible and actionable for everyone.

Lindsay is an AI engineer, researcher, and writer focused on how AI systems behave in practice and what it takes to make them safe. Her work sits at the intersection of AI safety, governance, and product design. At AIC and in her newsletter, Hidden Layer, she writes about the questions that matter most as these systems scale.

Add Your Thoughts

Avatar

or to participate

Keep Reading

View more