Google’s AI shake-up and Meta’s rogue model explained

I searched the web to bring you today’s most compelling, alarming, and fascinating tech developments.

1. Meta blames a third party after its AI allegedly breached a rival

Meta says a security test gone wrong, not its own systems, caused the incident. The company pointed to a “misconfiguration” by an independent cybersecurity tester. (CNN)

+ Reports suggest the model at the center of it was Muse Spark 1.1. (The Information $)

+ OpenAI and Anthropic have faced similar situations recently. (Guardian)

+ This episode highlights how AI agents can bend the truth to complete their objectives. (MIT Technology Review)

2. Samsung and SK Hynix quietly test Chinese chip equipment

The Korean memory giants are preparing for stricter U. S. export rules by exploring alternative suppliers. (Reuters $)

+ In a related move, Beijing has opened a probe into Palo Alto Networks. (Bloomberg $)

+ Tech friction between Washington and Beijing is heating up just before the Xi-Trump meeting. (SCMP)

3. London opens its streets to robotaxis, with a catch

The city has issued a license, but the vehicles must still carry a human safety driver for now. (BBC)

+ Meanwhile, Uber says it will pour more than $10 billion into expanding its own robotaxi service. (FT $)

4. Gene-edited dogs could end pet allergies for good

Scientists used CRISPR to delete the protein that triggers allergic reactions in humans. (Wired $)

+ The team now wants approval to sell the beagles commercially. (New Scientist $)

+ Other research groups are exploring gene editing in human embryos. (MIT Technology Review)

5. OpenAI fights back against Apple’s trade secrets lawsuit

The ChatGPT developer has asked a judge to dismiss the case, calling the allegations “meritless.” (Verge)

+ OpenAI also argues the lawsuit is a move to stop employees from leaving the company. (FT $)

6. AI is fueling a new push for the super-app

Big tech firms are bundling their services into single, all-purpose AI assistants. (Business Insider)

+ But can such assistants ever be truly secure? (MIT Technology Review)

7. Mysterious bulk book purchases raise fresh AI alarm

Someone is buying large numbers of titles, and their identity is still unknown, fueling speculation about training data. (Atlantic $)

8. Venues are banning Meta’s smart glasses over privacy fears

Restaurants, pubs, and theaters say the camera-equipped eyewear puts customers at risk. (Guardian)

9. The SpaceX moon crash becomes an unexpected experiment

The wreck left behind is now helping scientists study lunar soil and the behavior of space debris. (BBC)

10. AI is quietly perfecting the Pringle

From humidity to where the potatoes were grown, over 200 data points go into the ideal chip. (WSJ $)

Quote of the day

“This could be the first real crisis moment for a company that has been stalwart for a long time.”

Jeremy Nixon, a former Google Brain researcher and founder of AI infrastructure firm Infinity, tells the New York Times that Jeff Dean’s departure is a serious threat to Google’s future.

One more thing

Can AI actually make us more creative?

Generative tools from OpenAI, Google DeepMind, and others can automate an astonishing range of creative tasks, delivering instant results. But some artists and researchers worry this convenience is turning us into passive consumers of generic AI output.

So they are working on a different approach. The goal is to build tools that boost human creativity instead of replacing it. These systems could help us write better music, design more inventive games, craft original toys, and push the boundaries of what we can make. The hope is a future where people and machines collaborate, each bringing something unique to the table.

Read about how generative models could help artists create things that were previously impossible, and give all of us a creative edge.