OpenAI releases more realistic Images 2.5 models with deepfake safeguards
OpenAI released GPT-Image-2.5 Flare and Sunburst across ChatGPT, ChatGPT Work, Codex, and the API. The system card says greater realism can enable more convincing political, sexual, and sensitive deepfakes, while layered filters, C2PA metadata, and invisible watermarking mitigate misuse.
0 comments · 0 votes
Sign in to join the discussion →
No comments yet. Start the discussion.
Why it moved the index
The release combines more realistic, reference-faithful image generation with immediate distribution across all ChatGPT tiers and two hosted API models, increasing the scale and plausibility of synthetic-media misuse. OpenAI explicitly identifies more convincing deepfakes as a new safety challenge and documents prompt, input, and output checks plus provenance markers. Those controls materially reduce risk but do not negate the wider capability and deployment shift.
Assessment history
-
R1
Toward 53 · confidence 89
Adds the two exact Images 2.5 releases, their broad hosted access, deepfake risk, and documented safety stack.
10 Sept 2026
Share this page
-
DoomBench assesses “OpenAI releases more realistic Images 2.5 models with deepfake safeguards” as evidence moving toward doom, with magnitude 53 and confidence 89 out of 100 in the misuse and incidents category.
-
The DoomBench assessment of “OpenAI releases more realistic Images 2.5 models with deepfake safeguards” is based on reporting from OpenAI and records the editorial rationale, source quality, attribution, and revision history.
-
DoomBench summarizes “OpenAI releases more realistic Images 2.5 models with deepfake safeguards” as follows: OpenAI released GPT-Image-2.5 Flare and Sunburst across ChatGPT, ChatGPT Work, Codex, and the API. The system card says greater...
https://www.doombench.com/news/openai-releases-more-realistic-images-2-5-models-with-deepfake-safeguards-2026-09-08