OpenAI’s popular natural language model GPT-3 has a problem: It can be tricked into behaving badly by doing little more than telling it to ignore its previous orders.
Read full article on The Register
GPT-3 ‘prompt injection’ attack causes bad bot manners
- 7-Year-Old 0-Day in Microsoft Office Exploited to Drop Cobalt Strike
- AI Helps Improve About Managed Detection and Response
- TikTok, Flowmon, Cisco, Brokewell, RuggedCom, Deepfakes, Non-Competes, Aaran Leyland – SWN #381
- Impact of organizational structure on ransomware outcomes: Where does your org fit in?
- Hackers try to exploit WordPress plugin vulnerability that’s as severe as it gets