SS-NEWS-114: AI Cannot Be Trusted

Episode 114 June 05, 2023 00:49:21
SS-NEWS-114: AI Cannot Be Trusted
Security Serengeti
SS-NEWS-114: AI Cannot Be Trusted

Jun 05 2023 | 00:49:21

/

Show Notes

This week, we have three related AI articles all dealing with trust, or lack thereof, in the results AI so confidently gives you.  This can range from poisoning of the training data to confident hallucinations, to faking interviews.  It's a brave new world out there.

Article 1 - The poisoning of ChatGPT

Article 2 - Lawyer cited 6 fake cases made up by ChatGPT; judge calls it “unprecedented”
Supporting Articles:
Professor Flunks All His Students After ChatGPT Falsely Claims It Wrote Their Papers

Article 3 - Github - Ecoute

If you found this interesting or useful, please follow us on Twitter @serengetisec and subscribe and review on your favorite podcast app!

Other Episodes

Episode 87

November 21, 2022 00:31:34
Episode Cover

SS-NEWS-087: Interplanetary File System serving Malware from Mars

The Interplanetary File System serving malware from the stars!  Also included are some bonus discussions around automatically exfiltrating information from your own organization using...

Listen

Episode 132

December 18, 2023 00:55:16
Episode Cover

SS-NEWS-132: AI Drones, OAuth Abuse, and 23andMe!

This week we discuss Microsoft shutting down a bot network that created millions of fraudulent accounts, the coming AI Drone Overlords, OAuth Abuse, and...

Listen

Episode 8

May 09, 2021 00:52:17
Episode Cover

SS-NEWS-009: Phone Number Recycling, Malicious O365 Apps and Drone Hacking!

In this episode, we deep dive into two articles, with a bonus on hacking from drones where we wildly speculate on hacking attacks from...

Listen