Sora 2 Makes Videos So Believable, Reality Checks Are Required

According to Reality Defender, GenAI tools like Sora 2 are advancing faster than regulations can catch up, driving mounting concerns about deepfakes in enterprises. OpenAI’s Sora 2, launched in September with more advanced world simulation capabilities, can create eerily realistic videos from text prompts and images, and earlier guardrails have been tightened after SAG-AFTRA lodged a complaint.

Threat actors could abuse the technology to enhance social engineering, with experts warning that even PhD-level users struggle to distinguish real video from deepfakes with the naked eye. Defence strategies require a multi-pronged approach, including additional authentication factors such as likeness checks and device location, and random checks to thwart replication by adversaries.

Sora 2 updates also raise concerns about longer videos and improved voice authenticity, which could enable real-time manipulation during conversations. While watermarks offer some transparency, researchers doubt their effectiveness since threat actors may remove them, underscoring the need for ongoing verification of authenticity as the technology evolves.