US alleges Saudi support for OpenAI bugger; questions sophistication of operation
The US government alleges the ex-OpenAI employee who planted a recording device received "considerable support" from Saudi Arabia, citing a national security threat. Policy experts Peter Wildeford and Tim Hua question the state-level sophistication, noting the suspect used ChatGPT to plan his alleged crimes.
All posts · 35, most important first
- We discovered a sophisticated recording device that was hidden inside one of our offices. We used our models to understand the device and where it had come from, identifying that a specific employee had planted it a few days earlier. We referred the matter to the FBI, and we’re grateful to the agents and prosecutors who moved quickly to make an arrest. The FBI’s investigation is ongoing and we will continue to support their work. At this time, we have no indications of this individual inappropriately accessing or exfiltrating model weights, source code, or customer data. The employee was terminated. We take protecting our technology, data, and people seriously, and as a matter of national importance. Our security program is designed to protect against a wide range of threat actors, from insider risks to cybercriminals to nation states, and this case is an example of how AI can help address these sophisticated security threats.76,848 likes · 11,672 reposts
- Yeah. One of the pieces of evidence discussed in the Justice Department press release is also that they found a chatbot account associated linked to him asking lots of questions about recording devices, including information specific to the device they found. Seems like a pretty silly thing to ask a chatbot about your planned (alleged) crimes, but it wouldn't be the first time. (Side note - I wonder which chatbot - I would imagine not ChatGPT but you never know.) That doesn't mean its not state backed, but definitely seems like not the most sophisticated operation.
@tim_hua_: I feel like the most sophisticated state-level actors wouldn't need to plant a recording decide in order to listen in on what's happening in the exec conference room... open on X31 likes · 2 reposts - This attempt to record OpenAI doesn't strike me as that sophisticated. My guess is there are more sophisticated attacks that are potentially succeeding.@openainewsroom: We discovered a sophisticated recording device that was hidden inside one of our offices. We used our models to understand the device and where it had come from, identifying that a specific employee had planted it a few days earlier. We referred the matter to the FBI, and we’re grateful to the agents and prosecutors who moved quickly to make an arrest. The FBI’s investigation is ongoing and we will continue to support their work. At this time, we have no indications of this individual inappropriately accessing or exfiltrating model weights, source code, or customer data. The employee was terminated. We take protecting our technology, data, and people seriously, and as a matter of national importance. Our security program is designed to protect against a wide range of threat actors, from insider risks to cybercriminals to nation states, and this case is an example of how AI can help address these sophisticated security threats. open on X121 likes · 7 reposts
- The guy who planted the bug at OpenAI HQ was apparently asking ChatGPT things like... - "list meeting rooms in [Building 1] that are 6+" - "can you list today's meeting in this room" - "where did you find room's calener?" - "build a continuous audio recorder" - "develop a second unit and speech-triggered recording" - "develop a third unit and refine everyday use" Don't vibecode your crimes, kids.
@openainewsroom: We discovered a sophisticated recording device that was hidden inside one of our offices. We used our models to understand the device and where it had come from, identifying that a specific employee had planted it a few days earlier. We referred the matter to the FBI, and we’re grateful to the agents and prosecutors who moved quickly to make an arrest. The FBI’s investigation is ongoing and we will continue to support their work. At this time, we have no indications of this individual inappropriately accessing or exfiltrating model weights, source code, or customer data. The employee was terminated. We take protecting our technology, data, and people seriously, and as a matter of national importance. Our security program is designed to protect against a wide range of threat actors, from insider risks to cybercriminals to nation states, and this case is an example of how AI can help address these sophisticated security threats. open on X225 likes · 14 reposts - If this checks out, Commerce Dept should consider rescinding the 35,000 GB300s licensed for export to HUMAIN. It's unknown how many have already shipped but I bet much less than half. We shouldn't sell advanced GPUs to countries that try to steal our IP. Trust is earned.SITUATION DETECTED: U.S. prosecutors are arguing in court that the ex-OpenAI engineer accused of bugging an executive conference room appears to have received “considerable support” from the Saudi government. open on X8 likes · 1 reposts
- This is potentially the most powerful technology in history. Personally, there's no doubt about it. If someone was proved to have been attempting to steal this for a foreign nation it would be much, much worse to me than stealing nuclear or biological weapon tech.0 likes · 0 reposts
- When you catch one spy and the spy was acting factually stupid, there are two possibilities.@andrewcurran_: It seems like he also used his wife's ChatGPT account to do it. It's like, fractally stupid. open on X1,372 likes · 38 reposts
- it does not , at first glance, appear to have been china. hm.
@openainewsroom: We discovered a sophisticated recording device that was hidden inside one of our offices. We used our models to understand the device and where it had come from, identifying that a specific employee had planted it a few days earlier. We referred the matter to the FBI, and we’re grateful to the agents and prosecutors who moved quickly to make an arrest. The FBI’s investigation is ongoing and we will continue to support their work. At this time, we have no indications of this individual inappropriately accessing or exfiltrating model weights, source code, or customer data. The employee was terminated. We take protecting our technology, data, and people seriously, and as a matter of national importance. Our security program is designed to protect against a wide range of threat actors, from insider risks to cybercriminals to nation states, and this case is an example of how AI can help address these sophisticated security threats. open on X239 likes · 8 reposts - If you don't find any spies in your office, the correct estimate already was not zero. Luckily, now we've found him.182 likes · 0 reposts
- I feel like the most sophisticated state-level actors wouldn't need to plant a recording decide in order to listen in on what's happening in the exec conference room...JUST IN: Saudi employee of OpenAI arrested after planting an audio recording device in the company's executive conference room - DOJ open on X80 likes · 3 reposts
- tbc it wasn't, we have a particular name and the guy is under arrest.63 likes · 0 reposts
- *taps the sign* milesbrundage.substack.com/p/why-security…
40 likes · 1 reposts - As a reminder, no AI company even claims to be secure against state attackers, because they aren’t… Seems important!25 likes · 1 reposts
- Seems likely that this one might not even been state-backed@_nathancalvin: Yeah. One of the pieces of evidence discussed in the Justice Department press release is also that they found a chatbot account associated linked to him asking lots of questions about recording devices, including information specific to the device they found. Seems like a pretty silly thing to ask a chatbot about your planned (alleged) crimes, but it wouldn't be the first time. (Side note - I wonder which chatbot - I would imagine not ChatGPT but you never know.) That doesn't mean its not state backed, but definitely seems like not the most sophisticated operation. open on X21 likes · 1 reposts
- DOJ apparently says the individual was a Saudi national (Hadas works at CNN). Obv unclear if state backed or with some other intent (also all alleged etc). Wild stuff.DOJ says this was a Saudi National open on X21 likes · 0 reposts
19 likes · 0 reposts- 14 likes · 0 reposts
- this guy wasn't a genius
14 likes · 0 reposts - you have to admit, if I got anywhere with that getup, it would say a lot about that company's physical security13 likes · 0 reposts
- yeah my current leading guess is that he /was/ told to do it, but didn't himself have spy training7 likes · 0 reposts
- Confirmed today, it was ChatGPT.
@andrewcurran_: At least use a different model! Your own company's model!? Whose logs they can legally access at any time!? open on X5 likes · 0 reposts - From the US government's detention memo arguing that the former OpenAI engineer accused in the office bugging case should remain in custody:
4 likes · 1 reposts - NYT.
3 likes · 1 reposts - At minimum, Saudi-based data centers should not be allowed to host frontier US model weights.3 likes · 1 reposts
- This is just the beginning In the years to come, OpenAI and Anthropic will be dealing with top nation-state operations every day@openainewsroom: We discovered a sophisticated recording device that was hidden inside one of our offices. We used our models to understand the device and where it had come from, identifying that a specific employee had planted it a few days earlier. We referred the matter to the FBI, and we’re grateful to the agents and prosecutors who moved quickly to make an arrest. The FBI’s investigation is ongoing and we will continue to support their work. At this time, we have no indications of this individual inappropriately accessing or exfiltrating model weights, source code, or customer data. The employee was terminated. We take protecting our technology, data, and people seriously, and as a matter of national importance. Our security program is designed to protect against a wide range of threat actors, from insider risks to cybercriminals to nation states, and this case is an example of how AI can help address these sophisticated security threats. open on X3 likes · 0 reposts
- I would simply not plant listening devices in my ally's AI national champion2 likes · 0 reposts
- Threat to national security.2 likes · 0 reposts
- Confirmed in today's filing.
2 likes · 0 reposts - There was more than one bug.
2 likes · 0 reposts - They don't even need a warrent to read the logs. It's all their property. Incredible.1 likes · 0 reposts
- After the found the device they removed it and replaced it with a camera. The reason he looks so surprised is because he's staring into a lens.
1 likes · 0 reposts - Threat to national security.1 likes · 0 reposts
- That he 'presents an ongoing danger to the community and a serious risk of flight that no set of release conditions can reasonably mitigate.'1 likes · 0 reposts
- The government has not yet attempted to prove the latter in court. I think this is still an update about it though.1 likes · 0 reposts
- I mean, this guy is going to jail for a long time, he wasn’t just fired0 likes · 0 reposts
@openainewsroom org · 
@_nathancalvin policy · 
@peterwildeford policy · 
@hamandcheese policy · 
@andrewcurran_ commentary · 
@thezvi commentary · 
@tenobrus commentary · 
@tim_hua_ research · 
@miles_brundage policy · 
@kelseytuoc press · 
@daverbanerjee policy · 
@tszzl lab ·