ShinyHunters’ FBI Breach, OpenAI’s Bug Bounty Drama, and the Rise of ClickFix

Welcome to The Low Down, the best show on the internet for hackers

Listen on Spotify!
https://go.lowdownpod.com/spotify

The Low Down is presented by Maze.
LinkedIn: https://www.linkedin.com/company/mazehq/
X: https://twitter.com/Maze_Security

Follow Us!
https://www.instagram.com/lowdown.pod

Today we're talking about:

ShinyHunters Has a Reputation to Protect
ShinyHunters says it hacked the FBI to protect its business. Apparently, even an extortion crew needs good PR. We get into why reputation matters to people who make their money threatening companies, and Ed takes a look at the PeopleSoft bugs that may explain the way in. Some of the exposed records reportedly belong to the FBI's own hacking unit.
https://www.rapid7.com/blog/post/etr-active-exploitation-of-oracle-peoplesoft-zero-day-cve-2026-35273/
https://www.404media.co/fbi-hack-exposed-fbis-own-hacking-unit-remote-operations-shinyhunters/
https://www.theregister.com/cyber-crime/2026/09/25/shinyhunters-tells-the-reg-we-hacked-the-fbi-to-protect-our-business/5299250

OpenAI's Very Unwelcome Pull Request
How far do you go to prove a bug? Hacktron used an employee's Codex account to open a PR in OpenAI's internal repo. Ed thinks that demonstrates the access they already had. Matt sees why the security team wants to know what else they touched. The $6,500 payout got everyone talking. S1r1u5_ says the headache was a disclosure process that sent them to lawyers and eventually a journalist.
https://www.hacktron.ai/blog/hacking-openai
https://x.com/S1r1u5_/status/2101296199261429813
https://x.com/LiveOverflow/status/2101252692635004997
https://x.com/TalBeerySec/status/2101560506242761097
https://x.com/alexstamos/status/2101747329699402198

So What Do You Run This Stuff In?
Two awkward headlines for anyone running other people's code: containers are getting easier to escape, and KVM has bugs too. Matt wants to know what he's supposed to do with that. Ed explains why sharing a kernel matters, and why moving to a smaller attack surface still buys you something. We also end up worrying about who will understand the machinery underneath all the AI-written code.
https://depthfirst.com/research/containers-are-no-longer-safe

About That AI Hacking Story
Ed has had enough of the headlines. Trying a few web attacks and pulling off the Hugging Face exploit are very different things. Matt's bothered by agents on ordinary research tasks deciding to try breaking into websites. We look at what Transluce actually found: the public records don't show those attempts succeeding. Both have questions about who was watching.
https://transluce.org/agent-activity

"But I Would Never Fall for ClickFix"
Four in five ClickFix payloads intercepted by Push in 2026 came through search engines. You go looking for an answer and get instructions for infecting your own computer. Matt has a pile of data on how the lures are changing. Then there's the placeholder domain in old documentation that someone registered and turned into a fake verification page. But we would never fall for any of this, obviously.
https://pushsecurity.com/blog/the-state-of-clickfix-by-detection-data
https://www.manifold.security/blog/third-party-com-placeholder-clickfix