GPT-3, Generative AI, and cyberthreats

Over the last few months, AI has attracted even more attention than usual. Much of this is driven by OpenAI's ChatGPT tool, which allows anyone to create convincing, "human sounding" text from just a web browser.

But GPT-3 and generative AI can be misused, and could make it easier to carry out cybercrime or create fake news. Although ChatGPT has safeguards built in, the tools to create natural language text are becoming cheaper.

Security researchers at Finnish firm WithSecure put this to the test, in an EU supported project. They used a range of scenarios to see how "prompt engineering" could be misused, and how we can guard against it.

Our guest is WithSecure's intelligence researcher, Andy Patel. The full research report is also available for download.