Time Bandit ChatGPT jailbreak bypasses safeguards on sensitive topics

A researcher discovered a ChatGPT jailbreak named "Time Bandit" that exploits the model's "temporal confusion" to bypass safety filters and coax the model into providing detailed, often dangerous instructions (weapons, nuclear topics, malware). Independent testing by BleepingComputer and the CERT Coordination Center reproduced the issue and demonstrated the model producing actionable malware coding guidance; OpenAI has begun mitigations but the flaw persisted in some tests.