Hiding in Plain Sight: Prompt Injection Attacks Are a New AI Threat
Artificial intelligence is a massive productivity driver for every business category , including for law firms. Of course, as with any new technology , AI does not come without risks. That doesn’t mean that you shouldn’t use AI in your law practice . As an attorney, your obligation is not to eliminate risk, but to take reasonable precautions against it. The first step in that process is understanding which credible threats exist around the use of AI. Unfortunately, there is a new one you should be aware of: prompt injection attacks. Prompt injections are , in ess ence , commands hidden in AI data inputs . The AI tool will follow embedded commands, as well as direct user inputs, because it does not effectively differentiate between the two. Now, users can inject prompts on their own – some do so to try to ‘ jailbreak’ AI products – but, what business owners are more concerned about would be pr...