OpenAI talks slowdown

Good morning! In a world that has been dominated by how fast companies can develop AI, OpenAI is talking about hitting the brakes. 😳

[Sponsor FryAI]

This link leads to one of the most interesting things I’ve seen in AI recently.

What’s cookin’? OpenAI says AI has entered a new phase where policy needs to catch up fast, and CEO Sam Altman has even expressed that he’s willing to slow down if it means safer development. The company is calling for four big moves:

A major concern is AI helping build better AI. OpenAI says full “recursive self-improvement” is not happening yet, but AI is already speeding up research that once took human experts days. The company recognizes that labs need clear safety bars before pushing ahead.

🤔 Hunter’s take: The most interesting part is not the policy talk. It’s OpenAI CEO Sam Altman reportedly telling the team he is willing to slow down for the sake of safety. This is the same Sam Altman who has been willing to spend billions of dollars and stop at nothing to develop AGI. That matters because the AI race has mostly sounded like one giant foot on the gas. OpenAI itself saying, “There may be moments where speed is the danger,” is a pretty big shift. And it’s a huge red flag for everyone to pay attention to.

What’s cookin’? Anthropic says it blocked several attempts to use its Claude AI models in ways that could help develop biological weapons. The company’s new threat report says people also tried using Claude for cyberattacks, scams, propaganda, surveillance, and even weapons-related software. The tricky part is that biology is not simple to police. The same kind of information that could help someone make a vaccine or study a disease could also help someone with darker goals. Anthropic says it found five cases involving possible biological weapons support and shut them down, then used those findings to improve its safety systems.

🤔 Hunter’s take: I’m genuinely taken aback by this story. This is where AI safety stops sounding like sci-fi and starts sounding like a real public health concern. The danger is not that someone types, “help me destroy the world.” It is that powerful tools can answer small, technical questions that slowly remove barriers for malicious attacks.

Hi — I’m Hunter, a PhD candidate whose work has appeared in major academic journals and popular tech outlets. I founded FryAI to make staying ahead of AI clear, accessible, and fun.

How close are we to the point where AI surpasses human intelligence across nearly every domain and begins advancing beyond our control or prediction?