August 28, 2011 - Technology media outlet Tom's Hardware published a blog post yesterday (August 27) reporting thatinternetThe Engineering Task Force (IETF) recently released theAI Draft AI Content Disclosure Header Field.pendingweb page New machine-readable AI usage tags in HTTP responses.
The draft proposes to add a new HTTP header field to web server responses for marking AI involvement in web content generation, with the goal of enabling search engines, archiving systems, and browsers, among others, to quickly recognize AI-related content while keeping resource consumption low, 1AI cites a blog post describing the draft.
Unlike existing AI-generated content prompts (such as text statements or image/video watermarks), this HTTP header field is fully machine-readable without relying on human judgment or parsing complex metadata files, which officials say not only helps search engines index more accurately, but also improves the efficiency of compliance reviews and data archiving.
This header field will include five main pieces of information: mode (AI usage pattern), model (name of the AI model used), provider (model provider), reviewed-by (content reviewer), and date (date and time of generation or modification).
The mode contains the following four modes:
- none (no content generated or modified using AI)
- ai-modified (original human creation, modified by AI, e.g., spell-checking, summary generation, etc.)
- ai-originated (content generated by AI, modified and reviewed by human editors)
- machine-generated (almost entirely generated by AI with little human intervention).
The IETF notes that this standardized markup approach could reduce unnecessary computation and resource waste by providing a low-latency AI detection pathway for web crawlers, content analysis tools, and archiving systems, and that the draft, which has not yet become an official standard, may be further adapted and expanded in the future based on industry feedback.