Security researchers found ChatGPT-4o can be used for autonomous voice-based scams. Despite guardrails meant to provide safety, they found it easy to jailbreak. OpenAI’s safety team was disbanded earlier this year. - ThreadSky

brianpenny.bsky.social • 114 days ago

Security researchers found ChatGPT-4o can be used for autonomous voice-based scams. Despite guardrails meant to provide safety, they found it easy to jailbreak. OpenAI’s safety team was disbanded earlier this year.