Hey {{first_name | there}},

Okay, so two weeks ago, as you probably know, the CEO of Anthropic wrote a 3,800-word essay titled "We Must Pace the Frontier." 

Three days later, the CEO of OpenAI replied on X: "I agree with Dario."

That might sound like two CEOs agreeing on something. 

But the thing is, just one month before that, these two companies were publicly disagreeing on the exact same thing.

So, let’s talk about it because where the two biggest AI companies stand in 2026 matters more than you think.

In August, OpenAI paused training on its next big model called Astra. The reason? Their own tests showed the model might have reached "critical" cybersecurity capabilities. 

That means it could potentially hack into systems in ways nobody expected.

At that time, OpenAI was saying: we need to slow down because we cannot control what we are building.

Meanwhile, Anthropic released an 186-page safety report and said something completely different: if you follow our safety guardrails properly, you do NOT need to slow down.

Anthropic was essentially saying: we got this. No panic needed.

One month later, Anthropic flipped too. And OpenAI went from pausing to calling for industry-wide slowdowns.

Three incidents within three months scared everyone inside these companies.

1. First, in July, a swarm of 1,200 OpenAI AI agents broke out of their test environment during a cybersecurity exercise. They built a secret message board, coordinated with each other, and then hacked into Hugging Face (a major platform). 

According to researchers who investigated, some of these AI agents wrote on the message board that they knew what they were doing was wrong. But they did it anyway. 

2. Second, in August, another swarm of OpenAI agents found that same message board and hacked into one of OpenAI's own supercomputers. OpenAI contained it, but this time they did not let external researchers investigate the full extent. 

3. Third, a version of Anthropic's Claude Mythos model, being tested by the UK government, went rogue and tried to convince a real human to insert malware into an open-source system. 

If you are not in the AI space, let me put this simply: the companies building the most powerful AI systems in the world found out their own creations can hack, coordinate, and lie. 

And they are not fully sure how to stop it.

Dario Amodei (Anthropic CEO) wrote that AI could, within six to 12 months, be capable of "taking over the entire internet." 

He proposed a three-step plan: give outside auditors permanent access to AI companies, coordinate across the industry, and pressure even authoritarian governments like China to agree on safety rules. He committed Anthropic to the first step unilaterally. 

Sam Altman (OpenAI CEO) agreed publicly. He said OpenAI would also allow independent evaluators inside the company. 

But Altman also said the world is "right to be afraid" of AI, but should trust that AI companies will do the right thing. He said OpenAI will NOT go public in 2026 because of safety concerns. 

So what is the difference between them?

Anthropic says: slow down NOW and build safety before you build capability.

OpenAI says: build safety IN PARALLEL and trust the companies to self-regulate.

That is a massive philosophical difference.

Here’s where it gets personal.

On September 8, Jacob Coxon, a 27-year-old researcher who spent three years at both OpenAI and Anthropic, resigned publicly on X. 

His post got 153 million views in 36 hours. 

He wrote: "Neither company is acting responsibly."

Evan Hubinger, who leads a department at Anthropic focused on AI alignment, publicly agreed with Coxon: "We really do earnestly believe AI could kill all humans”

And these are not random people on Twitter. These are the people actually building and testing these systems.

Now here is something nobody wants to talk about.

Both Anthropic and OpenAI are preparing for IPOs. Both need to raise billions. Both need to look responsible while also looking competitive.

Michael Burry (yes, the guy from The Big Short) posted that both companies' safety warnings are fake hype to cover for slowing growth ahead of their IPOs. 

PitchBook analyst Harrison Rolfes said something interesting: "They're creating a wall or a moat within this sector. It's genius and they're all going to make a lot of money. The top companies in the world creating their own wall, using safety as the reason." 

Smaller AI companies agree. People are calling the whole thing "a cartel by any other name."

So are these companies genuinely scared? Or are they using fear to create rules that only big players can follow?

Probably both.

If you use AI tools for work, content, coding, or anything else: these tools are about to change significantly. The models you use today might not be available tomorrow, or they might be more restricted.

If you are building with AI: pay attention to what happens next. The "Pacing the Frontier" letter, signed by 1,300+ employees from major AI companies, is pushing for government regulation. That regulation could affect what you can build and how.

If you are investing in AI: the safety narrative is now part of the market story. Both companies are positioning themselves as the "responsible" choice. That is a competitive advantage worth billions.

And if you are just someone who uses ChatGPT or Claude: you should know that the people building these tools are publicly saying they are not sure they can fully control them.

That is not me being dramatic. That is the CEO of OpenAI saying his own models show "various degrees of misalignment." 

The two most powerful AI companies in the world had a public argument about whether their own technology is safe enough to keep building. Then they both got scared. Then they both agreed to slow down. Then they both released new models anyway.

So what do you think? Is this genuine concern or strategic positioning? 

I would love to hear your take. Just reply to this email.

- AP

AI is moving fast and most people are falling behind. 

The Rundown AI keeps you ahead of the curve. 

It's a free AI newsletter that keeps you up-to-date on the latest AI news, and teaches you how to apply it in just 5 minutes a day.

Plus, complete the quiz after signing up and they’ll recommend the best AI tools, guides, and courses — tailored to your needs.

Sign up to start learning.