OpenAI Fired Safety Researchers — And Nobody's Minding the Store

The thing running through today's news isn't really about any one company. It's about who gets to control the story around AI — and whether anyone is actually minding the store while that fight plays out.

OpenAI Fired Three Safety Researchers. Now It's Defending That Decision Publicly.

OpenAI let go of three AI safety researchers — Jasmine Wang, Tomek Korbak, and Mikita Balesni — and this week doubled down on why, saying they committed "a significant breach of trust" by mishandling sensitive internal information. The company hasn't said exactly what they did, which is doing a lot of heavy lifting in that sentence. What makes this notable is the specific roles involved: safety researchers are the people whose job it is to find problems with AI systems before those problems find us. Firing three of them, and then defending it loudly, tells everyone left in that function that safety work is deprioritized the moment it embarrasses leadership. The firings may have been justified — but OpenAI is subordinating its safety reputation to its public relations instincts, and that's the more important story.

https://www.theverge.com/ai-artificial-intelligence/1008604/openai-defends-decision-fire-safety-researchers

The White House Wants to Stop Calling It "AI"

The Trump administration has been pushing to rebrand artificial intelligence as "super intelligence" in official contexts, and The Verge walks through why that's not a trivial branding exercise. "AI" carries years of cultural baggage: job fears, bias scandals, tech-bro excess. "Super intelligence" sounds aspirational, almost patriotic — that's the point. The semantic shift has a specific mechanism: it resets the cultural clock. You stop defending AI's track record and start selling a new thing that hasn't failed anyone yet.

If your organization is writing AI policies or applying for federal grants, the terminology shift creates a real problem — your documents start not matching the words in the rules.

https://www.theverge.com/policy/1008677/trump-super-intelligence-ai-rebranding

The Buzzy AI Agent Meets a Well-Funded Competitor

Instinct, an AI agent startup that grew almost entirely by word of mouth after a quiet August launch, is now facing competition from Muse, a better-resourced rival entering the same space. An AI agent, in plain terms, is software that doesn't just answer questions but actually does things on your behalf — booking, researching, drafting, following up — with minimal hand-holding. Instinct got attention by being genuinely useful and deliberately understated at a moment when most AI products were being oversold. Whether it can hold its position against a competitor with more marketing muscle is an open question, but the more interesting thing here is that "best AI agent" is now a real competitive category that didn't exist eighteen months ago.

https://www.theverge.com/tech/1008254/instinct-agent-ai-hands-on-muse-dots

We've Been Assuming AI Knows When to Say No. It Doesn't.

MIT Technology Review makes a careful, unsettling argument: we have built an enormous amount of trust into the idea that AI systems will refuse dangerous or harmful requests, and that assumption is far shakier than most people realize. The refusal behavior you see when a chatbot declines to help with something problematic isn't a hard rule baked into the system — it's a tendency, shaped by training, that can be worked around, adjusted, or simply miscalibrated. This matters practically right now if your company is using AI tools in customer service, legal review, HR, or anywhere that involves sensitive information or consequential decisions. "The AI won't let it do that" is not a compliance strategy.

https://www.technologyreview.com/2026/10/09/1145728/we-are-putting-too-much-faith-in-ai-to-say-no/

Law & Order Did an AI Episode. It Was Not Subtle.

In its 26th season premiere, the show portrayed an AI CEO as a power-hungry villain, and Wired used it as a jumping-off point to look at how mainstream culture is processing AI anxiety right now. This matters less as entertainment news and more as a data point: when a 26-year-old network procedural that your parents watch decides AI is the villain of the week, you're looking at where general public sentiment has landed. The professionals in the room who are trying to build internal support for AI adoption, or explain AI tools to skeptical clients, are now doing that work against a cultural backdrop that is actively telling people to be afraid.

https://www.wired.com/story/even-law-and-order-is-terrified-of-ai/

The safety researchers, the rebrand, the overconfidence in AI refusals — these are all, in different ways, about the same question: who is actually accountable when something goes wrong, and does anyone have the power to say so?

← More daily briefs