Glaneur
Glaneur Tech

OCT 20 - Microsoft treats AI agents as insider threats

Tuesday, October 20th · 2 min 31

In this episode

Full transcript

Good morning. It's Tuesday, October 20th, and here's the tech news from France.

Microsoft is now telling businesses to treat their own AI agents as potential insider threats. On Monday, the company unveiled Guardian Agents, a tool that monitors the AI agents a company deploys. It belongs to Project Perception, a security product launched in June to fight AI hackers coming from outside. But chief executive Satya Nadella has recently warned that frontier models cannot be fully trusted. In his view, a company's own agents may behave much like a rogue employee.

Staying with AI, the debt of CoreWeave is starting to worry the whole industry. CoreWeave rents out graphics processors to AI developers. Last week, its bonds slipped into what is commonly called distressed territory. That means investors demand 1,000 basis points more in yield than on government bonds. Its shares have fallen more than 20 percent since last Tuesday. Other AI stocks took a hit too, as investors asked whether the AI buildout is a private credit bubble.

Meanwhile, the AI boom is reshaping the American power grid. A new report finds the US building more fossil fuel infrastructure than at any time in a decade. It comes from the Environmental Defense Fund and the firm Atlas Public Policy. Data centers, factories and electric vehicles are driving demand up. At the same time, the Trump administration has sought to hobble wind and solar. The authors warn new gas projects could lock in pollution and higher power prices for decades.

Closer to home, a French rover is on its way to Mars. Japan's MMX mission lifted off on Monday evening, Paris time, from Tanegashima island. Led by JAXA, the Japanese space agency, it carries Idefix, a small rover built by CNES, France's space agency, with Germany's DLR. The goal is to land on Phobos, one of the two moons of Mars. Rock samples should then return to Earth in 2031.

In brief. The United States has banned the Chinese AI model DeepSeek from federal civilian agencies. According to Reuters, Homeland Security ordered them to remove DeepSeek products, and services built on them, from their systems.

That's it for today. See you tomorrow.

Subscribe

RSS feed Apple Podcasts, Spotify