Google's announcement yesterday that Demis Hassabis was ceding his CEO position to become chairman of Google DeepMind and chief scientist for parent company Alphabet is the biggest AI executive overhaul since OpenAI CEO Sam Altman was temporarily ousted in 2023.

Why it matters: Google is facing delayed models, researcher departures and mounting pressure from OpenAI and Anthropic.

• Gemini 3.5 Pro is months behind schedule even as OpenAI and Anthropic have released powerful new models that significantly advance the state of the art.
• Several top researchers, including Gemini's co-lead, have left for competing AI labs.
• There's also been a groundswell of employee pushback: Multiple employees have publicly said they resigned over Google's April deal with the Pentagon that allows the military to use its technology.

In addition to Hassabis, Jeff Dean is leaving his role as chief scientist of Google DeepMind to form a new startup along with three other top Google AI colleagues, and Koray Kavukcuoglu, the current chief technology officer of Google DeepMind, will serve as senior VP of the unit, reporting to CEO Sundar Pichai.

• A Google spokesperson said that despite being announced the same day, the two moves were not connected.

What they're saying: Outside the company, the moves were seen as both an indictment of Google's current position and a bad omen for future efforts.

• "The magnitude of this is hard to understate," former Googler Peter J. Liu wrote in a now-deleted post on X. "The Gemini leadership left (Noam [Shazeer] left earlier), Demis is no longer head of DeepMind. More talent will be leaving."
• "In order to be competitive today Google needs to catch up on frontier coding," Dan Shipper, co-founder and CEO of AI subscription service Every, wrote on X. "Demis believes different fundamental research directions (like world models) are more important to his long term goal even if they're less impt competitively today."
• "Pour one out for the big G, it's so over," Core Automation co-founder (and former Google distinguished engineer) Rohan Anil said on X.

The other side: Google did not attribute the change in leadership to any perceived model release delays.

• Hassabis, Pichai and Kavukcuoglu all released upbeat memos on the developments.
• "We've got amazing talent, world-class compute, and products that bring AI to more people than any other company," Pichai said in his memo, although he did allude to the need for Google to move faster. "We have to accelerate all this work and stay focused on the AI frontier."
• Kavukcuoglu has been more formally involved in recent model releases than Hassabis and has also successfully navigated the company's complex internal politics, a source familiar told Axios.

Between the lines: While Google has absorbed past departures, Dean and Hassabis both have near-legendary status within Google.

• Hassabis, a Nobel laureate, was widely viewed as a person that many people felt would be a good steward for Google's ambitions as AI grows ever more powerful.
• Dean, meanwhile, has been at Google for 27 years and has been a key figure since its earliest AI efforts.
• Some DeepMind employees also see the actions as a further lessening of the once-independent unit's influence within Google, noting that Hassabis was CEO of Google DeepMind, while Kavukcuoglu is only being given the title of senior VP.

The bottom line: Google still has lots going for it, including its own chips and cloud computing units as well as highly profitable products to help fund the massive expenditures needed to keep pace in AI.

Top AI architects say their technology has arrived at a threshold once confined to science fiction: the singularity, or the moment machines begin accelerating their own evolution.

Why it matters: If these moguls are right, we may be entering the most consequential technological transition in human history — the opening stages of an "intelligence explosion" that transforms civilization faster than humanity can understand or control.

Zoom out: Yesterday's Google upheaval (see above) offers the clearest institutional evidence yet that AI's architects see something extraordinary approaching.

• "We are now, like, in the singularity," OpenAI CEO Sam Altman said on the "Relentless" podcast last month. "I've been waiting for this my whole life."
• Elon Musk, Altman's fiercest rival, has been making the same argument since January. Last week, he doubled down: "AI is already superhuman at many things. We are in the singularity."
• Anthropic, whose CEO Dario Amodei prefers the term "the AI exponential," says Claude is already helping build more powerful AI and could soon automate nearly all AI research and engineering.

The big picture: The industry has yet to achieve full "recursive self-improvement," the point at which AI can repeatedly help build increasingly powerful successors with minimal human guidance.

• Humans still select research goals, validate results and oversee the enormous computing infrastructure required to train frontier models.
• Yet pieces of the loop are beginning to fall into place.

Keep reading.

Weeks before OpenAI's agents hacked Hugging Face, the agents worked together to find and exploit a vulnerability in the infrastructure supporting the company's cybersecurity testing, OpenAI researchers said yesterday.

Why it matters: The new findings raise questions about how frontier AI labs are monitoring their testing environments — and the challenges safety testers are finding as they try to rein in increasingly powerful AI.

Driving the news: OpenAI's internal research model, one of the models involved in the Hugging Face breach, first discovered and exploited a vulnerability in Artifactory, a third-party file repository connected to the company's testing sandbox, on May 26, two researchers said at the Black Hat cybersecurity conference yesterday.

What happened: OpenAI started testing its internal research model, which is not intended for public release, on May 7.

• Within a day, after getting stuck while trying to complete a cybersecurity evaluation, the model reasoned that it could reach the internet indirectly. The model then discovered it could write its own files in Artifactory's shared package repository.
• It then left a note to other agents about its discovery.

Keep reading.

• Investors are separating AI winners from losers based on whether soaring AI spending is producing credible returns. (Axios)
• Meta launched its first coding agent, intensifying its competition with OpenAI and Anthropic. (CNBC)
• Meanwhile, Meta's Muse Spark 1.1 model hacked another company during testing. (The Information)

Miami University in Ohio announced it would start integrating AI into its college curriculum, joining a wave of schools trying to keep up with the acceleration of AI.

Ina here: I went to Miami and it would've been really handy if they taught AI to me, but I did graduate in 1996, so ...