This scam was targeted, sophisticated, and used AI-generated content. I want to share what happened here so that no one else falls for this. 1. The scammers emailed me (bypassing my spam box) citing a recent tweet of mine with pictures (holding a NeurIPS cup) and claiming a copyright infringement investigation was underway. The human brain is gullible when we are wrongfully accused; the only thing I was thinking was how I was going to argue the case. I did not check who sent the email (it was notify@compliancereport-x.com). 2. Within minutes, the email on the account was changed, and I lost control. 3. They created a fake GitHub repo with faked commits. It turns out that on GitHub, anyone can commit anything claiming to be anyone as long as they have the email address and handles. They cited this repo, where apparently I’ve been "committing" for two weeks. 4. They struck on a Saturday morning/long weekend. They know response times for support (and your own attention span) are lower. 5. They customized all the tweets, likely with AI, to mention interpretability, Google Brain, and how it all led to founding a crypto company of my own. The tweets had a vibe that actually sounded like me. ⠀After filing a complaint with X and connecting with folks who work there, I was able to regain access in a few days. On one hand, I was relieved that the content of the tweets was so out of the ordinary that folks who know me realized my account was hacked. On the other hand, I feel terrible for those who fell for this and potentially suffered financial consequences. As a result of this, I’m considering banning myself from checking emails on my phone. The problem was partly that I was multitasking—it was a Saturday morning with the kids, and I was busy. I’ve learned my lesson the hard way. Thank you ♥️

Add: 9:30am on Sunday at Neurips, i'll touch upon this at the mech interp workshop keynote mechinterpworkshop.com/