Open Models Will Do Anything: AI Is Speeding Up Attacks | Rob Chapman, ArmorCode
Rob Chapman of ArmorCode took the long way into security, from bulletin boards in middle Tennessee to running security for oil and gas companies. He says anyone claiming AI is not speeding up attacks is not paying attention. We got into open models that push back once and then go along, exploits weaponized in minutes, why most exploitation still comes through features and bad config, and why 'patch everything in two weeks' is not a program.
Key takeaways
- Stock open models can be 'good enough' to do anything meaningfully dangerous, even if they are not the frontier.
- The gap between a disclosure and a working attack is now measured in minutes, not weeks.
- Most exploitation still comes through features and bad configuration, not only through CVEs.
- 'Patch everything in two weeks' is a slogan unless there is a program and an army of automation behind it.
- A year ago a lot of teams did not want AI touching remediation. Now a lot of them do, which makes the permission question urgent.
Transcript
Are they as strong as the latest and greatest frontier? No, there's some research showing that they're still leading edge there, but it's certainly good enough to do anything meaningfully dangerous.
Rob Chapman, 20:35
More like this
Don't miss an episode
New conversations, in your inbox.