Responsible Scaling Policy

Anthropic's public safety policy that specifies in advance the safeguards required when model capabilities cross defined thresholds.

2 articlesLast mentioned

The Responsible Scaling Policy (RSP) is a safety policy first published by the AI company Anthropic in September 2023. It commits the company not to train or deploy models that cross predefined capability thresholds until corresponding security and deployment safeguards are in place, and it defines a tiered system of AI Safety Levels (ASL) whose requirements become stricter as model capabilities increase.

The ASL framework is modeled on the biosafety levels used in biological laboratories. The policy sets capability thresholds for risk areas such as assistance with biological and chemical weapons, offensive cyber capabilities, and the automation of AI research and development, and it specifies the evaluations used to detect those thresholds and the measures triggered when they are reached. Substantially revised in 2024 and updated since, it is regarded, alongside OpenAI's Preparedness Framework and Google DeepMind's Frontier Safety Framework, as one of the principal safety frameworks adopted by frontier AI developers.

This entry is based on AIPOST articles and widely known facts. If something is wrong, please send us a correction request.

Articles covering this entry

  • METR Audit Finds Anthropic AI R&D Running at Twice Its Historical PaceAI News ·

    …in the September pledges, and the engagement was separate from the independent evaluator framework in Anthropic's October 15 Responsible Scaling Policy update.

    Read article
  • Anthropic S-1 RSI Disclosure: 2x Research Speedup, Not a Runaway LoopAI News ·

    …nternal estimates indicate the company may imminently meet its own Responsible Scaling Policy definition of recursive self-improvement (RSI), the point at which…

    Read article

Found a mistake or missing an entry?

Wiki entries are based on AIPOST articles and widely known facts. If an entry is wrong, send us a correction request and we will check and fix it. Names from new articles are added to the wiki after the article is published.