Why Google’s AI staff are challenging contracts

Google AI researchers want the right to speak

Google researchers say their contracts could stop them from warning the public about AI risks, even after they leave the company.

According to The Washington Post, employees in Google's AI division are pushing executives to remove non-disparagement restrictions from their employment agreements. The dispute became public on November 3.

The concern is serious. Researchers working directly with advanced AI systems may notice problems that people outside Google cannot see, and contractual restrictions could discourage them from discussing those risks openly.

I think this raises a reasonable question about how much control AI companies should have over what their own researchers can say. A warning shared after someone closes their work laptop could be valuable to regulators, independent scientists, and the public.

Other AI companies may face similar pressure to clarify what employees can disclose.

If researchers believe the public needs to know something, should their employer get to decide whether they can say it?

Apple wants your watch to remember conversations for you

Apple's latest Watch feature can remember what someone said 15 seconds ago, even when you weren't paying attention.

The company has started beta testing four AI-powered audio features for Apple Watch, including Live Rewind and Siri Recaps. Users can revisit missed words on their watch screen or get short summaries of conversations, including meetings and everyday discussions.

That sounds useful. Especially when someone mentions a deadline during a busy meeting and you don't have a pen nearby.

But I think the bigger issue is consent. Apple says the features are optional, audio isn't permanently stored, and conversation recaps are encrypted, yet people speaking to a Watch wearer may not realize their words are being processed.

If this becomes a standard smartwatch feature, people may start expecting conversations to be remembered automatically, even when nobody explicitly agrees to it.

Will Apple's privacy safeguards be enough to make people comfortable talking around someone wearing one?

OpenAI is testing how much engineering AI can handle

OpenAI used GPT-6 Astra to design a physical device, from its circuit board to its outer casing, with no direct human edits.

Two employees completed the design in under a week, and a manufacturer produced prototypes within a month. OpenAI is also working with Mazda on vehicle aerodynamics and Chip Ganassi Racing on converting 3D scans into manufacturing designs.

The interesting part is how much work AI can already handle inside engineering software. I think this could be especially useful for teams spending hours preparing CAD drawings, running simulations, and revising parts before manufacturing.

But the prototype still needed several rounds of human review because Astra missed manufacturing tolerances and assembly requirements. Those details matter when a design has to become a working physical product.

If these systems improve, engineers could spend less time on repetitive design tasks and more time checking whether the results actually work.

How much of the engineering process will companies eventually trust AI to handle?

Prohuman team

Free Guides

Explore our free guides and products to get into AI and master it.

All of them are free to access and would stay free for you.

Feeling generous?

You know someone who loves breakthroughs as much as you do.

Share The Prohuman it’s how smart people stay one update ahead.