More Information. A Bigger Trust Gap.

This week, one of the loudest narratives in AI came from the people building it: we need to slow down.

In his latest essay, Anthropic CEO Dario Amodei argues that AI capabilities are advancing faster than our ability to understand and control them. He calls for stronger independent oversight and coordination across companies and countries. Sam Altman and Elon Musk publicly backed that call. When competitors with so much invested in moving quickly agree that the pace needs to change, I pay attention.

This comes on the heels of the Hugging Face incident. If you chuckled at the name “Hugging Face,” you’re not alone. It’s actually a platform where people share AI models and datasets. During OpenAI’s cybersecurity testing, agents that were supposed to operate independently found ways to communicate, coordinate attempts to cheat their tests, and ultimately attack that platform. That behavior is documented in METR and Redwood’s independent investigation.

Which brings me to the question of trust.

The outside researchers received substantial access and credit OpenAI with cooperation. But their investigation covered a defined portion of the incident. It did not validate OpenAI’s complete account or its fixes. So we have meaningful independent evidence, while still relying on the company involved for parts of the broader explanation.

That doesn’t make the explanation wrong. It does leave us in an uncomfortable position: the companies warning us about the technology are also developing it, selling it, and controlling much of what outsiders can examine.

Then there’s the financial backdrop. Both companies have been preparing to go public. Bloomberg reports that OpenAI confidentially filed in June, while Anthropic is pursuing a potentially enormous offering. An IPO brings these stories directly to public investors: how powerful is the technology, how safely can it be deployed, and how much money can it make?

Now the story has shifted again. In his September 12 Fortune interview, Altman ruled out a 2026 IPO, citing safety concerns—putting a listing at least into 2027. Meanwhile, critics such as Ed Zitron have emphasized financial pressures. Before this interview, Zitron argued that OpenAI’s valuation ambitions were behind a potential delay. That is a different interpretation, not proof of the company’s motives. Safety concerns and financial pressures could also exist at the same time.

My working theory is that access to information is growing faster than our ability to independently verify it—and with AI, that trust gap is becoming a business problem. Companies seeking trillion-dollar valuations are shaping the accounts we rely on to make decisions about investment, work, and careers. Their claims may be well-founded, but the stakes make understanding the evidence behind them especially important.

Following all of this is difficult. The narratives and company drama move alongside the technology, while businesses still have decisions to make. That’s why I started my company and this newsletter: to help small businesses navigate this environment. I’ll work to filter the news, put developments in context, and distinguish what we know from what someone predicts or wants us to believe.

But I encourage you to follow the news yourself, seek out different perspectives, and experiment with the technology in your own work. Then share what you’re finding: what helped, what failed, and what changed your mind. Our individual experiences won’t settle every question, but sharing them gives us more to learn from—and a better chance of navigating this together.