The Wall Street Journal reports on how “indirect prompt injection” can lead to safety failures in large language models without proper safeguards. It comes as OpenAI CEO Sam Altman warns that consumers are at greater risk of fraud as bad actors use AI to gain access to accounts.
Full story