Last week a researcher quit his job at Anthropic, walked away from about $2 million in stock he had not yet been paid, and wrote that the people building AI believe it could kill us all by the end of the decade.

153 million people read it in 36 hours. Four days later his own chief executive published an essay saying the whole industry should slow down, and the bosses of OpenAI, Google and Meta agreed within a day.

Then on Monday, a company raised $40 million to check AI and sell insurance on the answer. That last part is the one worth paying attention to.

Let's get into it.

Found the right person, but they’re in a country where you don’t have an entity? Setting one up can take months and significant cost.

Remote removes that barrier by becoming the legal employer through our own entities — handling compliant contracts, local benefits, tax setup, and onboarding for you. In fact, an employee is onboarded to Remote every 7 minutes.

Once they’re hired, the same in-house teams that support employment locally also run payroll — so you’re not bouncing between disconnected providers. Less setup, less complexity, and less time between finding the right person and getting them started.

See how it works

On September 8, a researcher named Jacob Coxon quit Anthropic and posted that the labs are racing straight to self-improving superintelligence and gambling with our lives. He gave up roughly $2 million in unvested stock to say it out loud.

The post was seen 153 million times in 36 hours. His colleague Evan Hubinger, who runs alignment research at Anthropic, backed him up and put the odds of AI killing everyone above one in ten this decade.

Four days later, Anthropic's chief executive Dario Amodei published an essay arguing the industry should deliberately slow down and let outsiders inside the labs to check the work. Within 24 hours the bosses of OpenAI, Google DeepMind and Meta all said they agreed.

Sam Altman also cancelled OpenAI's stock market listing for the year, calling it an ill-advised moment to go public. Take everybody's motives at whatever discount you like.

Here is the part that actually changes something. When an industry says out loud that it needs checking, somebody has to sell the checking.

On Monday a company called AIUC raised $40 million to do exactly that, which takes it to $55 million raised in total.

What it sells is simple enough to explain in a sentence. It runs an AI system through 5,000 different attacks and failures, gives a stamp to the ones that pass, makes them retake the test every three months, and then sells insurance against the failures its own test is looking for.

Cursor, ElevenLabs, Harvey, Lovable, UiPath and KPMG already carry the stamp. The founders say they are copying Underwriters Laboratories, the safety testing company set up in 1894 because insurers refused to cover buildings wired for electricity until somebody would certify the wiring.

It is not one company having a clever idea. There are now more than 400 businesses selling some version of checking AI, holding about $12.5 billion between them, and well over half were started in the last three years.

We have watched this happen before, and the early forecasts always come in low. When Europe brought in its privacy law in 2018, the estimate was that the world would need about 75,000 privacy officers, and within a year roughly half a million organisations in Europe alone had appointed one.

Why it matters: industries like this get built in the gap between everyone agreeing something needs checking and anyone agreeing how to check it. That gap is wide open right now, which is why the standard, the inspector and the insurance company for AI are currently all the same startup.

Almost all of that $12.5 billion went into building products. Somebody still has to run the actual assessment, write the policy, gather the evidence and sit in the room with the client.

That is a services business. It needs no funding to start, and last time around it ended up much bigger than the software it grew next to.

Watch who is moving and the shape is obvious. KPMG launched its own AI assurance practice in September 2025, and KPMG also now carries AIUC's stamp as a paying customer.

That is the tell. The big firms are buying this and selling it at the same time, which is exactly what happened when security audits and privacy compliance first became billable work.

The rulebooks to build a practice on are already public and free to read. AIUC-1 covers AI agents, ISO 42001 covers how a company manages AI, the NIST AI Risk Management Framework is the American reference, and Europe's new transparency rules took effect in August.

Certificates are the cheap way to look credible while the market has no established ones. ISACA's AI audit exam costs $459 for members and $599 otherwise, but only if you already hold an accounting or audit qualification. The IAPP's AI governance exam costs $649 and $799, has no entry requirements at all, and you can sit it from home any time in the next year.

Why it matters: the product side of this market is already crowded and well funded. The advisory work around it is neither, and that is historically the part that absorbs far more people and far more billing than the tools ever do.

Land your next role faster with Tailr, our Claude-powered job search and networking tool: Start searching

Find the assurance gap in a market you already know

Every new technology creates a market for vouching for it, and the opening is usually visible before anyone has named it. This prompt hunts for that opening in an industry you already understand, which is the only place you have an unfair advantage. Paste it into Claude or ChatGPT:

Step five is the one to take seriously. If the objection it raises is one you cannot answer out of what you already know, that is useful information for the price of five minutes.

Three polls settled this week, and all three went the way the consensus expected.

For calibration, the last twelve months averaged about 31,000 a month, and August came in at 162,000 against a consensus near 56,000. Reply and tell us your pick.