OpenAI is adding text watermarking in ChatGPT and Codex

OpenAI has begun deploying an invisible, machine-readable watermarking system for text outputs across ChatGPT and Codex. This new approach aims to embed subtle markers within AI-generated text, making it easily identifiable by automated systems while remaining imperceptible to human readers.

In this initial rollout phase, the watermarking feature is being introduced exclusively for users located within the European Union. This phased geographic approach aligns with the stringent regulatory frameworks and compliance standards operating in the region.

According to OpenAI, its proprietary "textGrain" watermarking method matched or exceeded the performance of other existing approaches, such as Google DeepMind's SynthID for text. Notably, Anthropic also announced a similar watermarking mechanism back in August, highlighting an industry-wide push for AI transparency.

The introduction of reliable text watermarks is a crucial development in addressing misinformation, deepfakes, and academic integrity issues associated with generative AI. Establishing a verifiable origin for text content helps build greater trust in digital communications.

For the global tech ecosystem, including emerging markets, these updates signify a major shift toward standardized AI accountability. As these watermarking technologies mature and expand beyond the EU, developers and platforms worldwide will need to adapt to the new paradigm of AI-generated content tracking.