The FTC's inquiry also covers concerns about the use of large language models, such as those used by OpenAI and competitors like Google's Bard. These models are being used by banks for tasks like organizing institutional knowledge and providing customer service via chatbots. The FTC has requested information from OpenAI about a data breach in March where some users could see other users' payment-related information, and about "prompt injection" attacks that can cause the model to output information it has been trained not to provide.
Key takeaways:
- The Federal Trade Commission is investigating OpenAI for potential 'unfair or deceptive privacy or data security practices'.
- The FTC has requested details from OpenAI about a data breach in March where some users could see other users' payment-related information.
- OpenAI is also being asked to share information about 'prompt injection' attacks that can cause the model to output information it has been trained not to provide.
- Banks that have launched AI chatbots have been careful to limit their capabilities to avoid giving out incorrect or sensitive information.