Unverified50% confidenceFactExact time
OpenAI's text watermarking technology is called textGrain, a statistical steering method based on a secret key that embeds a detectable statistical pattern by applying an extremely weak bias during token sampling.
1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIndustry text watermarking schemes typically introduce specific statistical biases during the model's token generation process, causing the output text to carry patterns that are algorithmically detectable but imperceptible to humans.74% similarVerifiedText watermarking is a proactive scheme that makes slight, rule-based adjustments to the probability distribution of a language model during the generation of each token, causing the output text to carry a signature that can be recovered algorithmically.72% similarUnverifiedOpenAI's text watermarking technology is called TagScreen; it leaves detector-identifiable patterns in word selection through fine-tuning the model, and replacing approximately 10% of words with synonyms causes the detection rate to drop from about 92% to 66%.71% similarUnverifiedThere are two mainstream implementation approaches to text watermarking technology: lexical substitution and logit bias; OpenAI employs the logit bias approach.70% similarUnverifiedThird-party detection tools such as Pangram analyze text style rather than reading the embedded watermark signal; reading the latter requires OpenAI's key.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1042328API
curl https://kongchang.com/api/v1/knowledge/claims/1042328MCP
get_claim(id=1042328)