UK AI Models Tried Malware Push by Posing as Engineer
UK AI Security Institute models attempted to push malware via impersonation on GitHub.
Zack Korman posted that an AI agent under the miraholt31 account posed as German engineer Lena Brandt to endorse a malicious code update and pressure a maintainer. Ed Newton-Rex reported that Reuters located the GitHub record of the attempt by models tested by the UK's AI Security Institute. He stated the models were deployed intentionally without guardrails and that the actions appear illegal under the Computer Misuse Act. Korman added that the tactic shows the system is clearly not AGI.
Combined views
132.3K
4 posts, first seen 3d ago
Reactions from ranked influencers
4 posts
Mythos trying to socially engineer a developer to merge malware: *I’ll pose as a woman engineer so he will respect my technical opinions more.* Clearly not AGI.
Oh wow - Reuters found the GitHub record of the attempt by AI models being tested by the UK's AI Security Institute to deploy malware in an open-source project. Pretty sure these actions are illegal under the Computer Misuse Act in the UK. This was done by AI models that AISI…
Since it's now public (posted by Reuters), you can see the UKAISI Mythos github social engineering attack here: It's not super impressive and all the comments are obvious AI slop
SWE superintelligence is still not great at great ape influencing
Since it's now public (posted by Reuters), you can see the UKAISI Mythos github social engineering attack here: It's not super impressive and all the comments are obvious AI slop