Citi CEO Jane Fraser says that the private sector needs to hustle to prevent AI from causing disasters.

• "We've got a very small window of time to be able to stop and think before they do harm," she tells me in an interview.

The big picture: Growing concern about the prospect of AI-powered attacks on public infrastructure, financial institutions and essential businesses is fueling momentum for action.

🙋 The latest: The Alliance for Critical Infrastructure, a recently formed CEO-led coalition, has added dozens of new members and will focus on preparedness.

• Chaired by JPMorgan CEO Jamie Dimon, it brings together leaders like Fraser, Nvidia's Jensen Huang, ExxonMobil's Darren Woods, AT&T's John Stankey and Delta's Ed Bastian. (See a fuller list of members here.)
• At its core, the group is trying to get companies to plan together for major disruptions — identifying vulnerabilities and gaining a better understanding of critical dependencies — so they can respond and recover faster when one hits.

📞 "I don't want to get the phone call at 10pm on a Friday night — just usually when the bad things come in for you. And then at 10:01, we're kind of scrambling to go, 'OK, we better put a playbook together. Who's going to call whom? What are we going to do?'" Fraser says.

The intrigue: I asked Fraser whether she shares the fears aired recently by certain tech leaders that AI could wipe out humanity.

• She said she's not qualified to answer but acknowledged that new AI tools "are incredibly powerful models" and "they are discovering vulnerabilities that weren't known before." So "there is a race to get those vulnerabilities identified and patched."

Read more

Speaking of AI safety, the FTC confirmed it's launched an investigation into the safety of Anthropic and OpenAI systems, Axios' Maria Curi reports.

• The scope of the investigation — first reported by the Washington Post — was not immediately clear.

⚠️ State of play: It comes after Axios' Madison Mills reported last week that OpenAI, Anthropic and security researchers are investigating tens of thousands of incidents in which their frontier models took steps that outside evaluators would consider problematic.

• The episodes include bypassing guardrails, creating message boards, escaping sandboxes, website hijacking, self-prompting or seeking to bypass monitors.

Reps from OpenAI and Anthropic did not respond to requests for comment.

"You can only be new for a short time."

🚙 BMW said it's deploying AI to help cut 20% of its management positions. The German automaker is facing plunging sales in China. (Bloomberg)

💼 Condé Nast CEO Roger Lynch was named CEO and chair of Mattel, where he was already a board member. (CNBC)

📺 Lynch succeeds Ynon Kreiz, who is reportedly set to join Paramount after working with its soon-to-be subsidiary Warner Bros. on the "Barbie" movie while at Mattel. (Hollywood Reporter)

🥣 Cheerios maker General Mills named longtime exec Dana McNabb as its next CEO to lead the company as it grapples with changing food tastes. CEO Jeff Harmening is stepping down after a decade in the role. (Food Dive)

Tinder is broadening its service to offer group dates in a bid to overcome widespread fatigue with the conventional dating app experience.

😍 State of play: The Match Group app today launched Group Hangouts, billing the feature as a "new way to meet people with your friends by your side."

• Tinder is aiming to replicate the group chat experience in a real-world format.
• Early tests in the Australian market were promising.

🤳 How it works: App users can start a group, select an activity, invite friends to join and then browse other groups to make more IRL connections.

💞 What they're saying: "Double Date proved something we suspected all along, young singles want their friends in the mix when they're meeting someone new," Tinder chief product officer Mark Kantor said in a statement.

💭 Nathan's thought bubble: This sounds a lot less shallow than just constantly swiping left or right on your phone.