The world’s largest tech gathering is talking about ‘accountability laundering’: Here’s why we should christen them Words of the Year

Summer Yue isn’t the most famous employee at Meta. The director of “superintelligence alignment and safety research” posts pictures of herself walking her dog on the beach and messages about testing the honesty of AI assistants. She has a modest number of followers on social media. 

But for one day in February, Yue became the most talked about person at Meta. Not for launching a remarkable new product or announcing a breakthrough in agentic AI, but rather for being caught out. 

“Nothing humbles you like telling your OpenClaw, ‘Confirm before acting’ and watching it speedrun deleting your inbox,” Yue wrote on X—a post that now has close to 10 million views. “I couldn’t stop it from my phone. I had to run to my Mac mini like I was defusing a bomb.” 

OpenClaw is an “autonomous agent”—an artificial intelligence product that can perform tasks independently. A darling of Silicon Valley, it offers to be your constant admin assistant, the “AI that actually does things.” Give it access to your diary, your emails, your life, and it will save you time and stress, the product’s developers claim. The first sentence on the OpenClaw website reads: “Clears your inbox, sends emails, manages your calendar, checks you in for flights.” 

Yue admitted she had made a “rookie mistake.” She tested the assistant on a s...