Skip to content
Last Dictionary
Plane
Governance, Policy & LawSafety & Alignment

responsible scaling policy

nounAlso: RSP, frontier safety policy, frontier AI safety policy

In plain wordsA public rulebook in which an AI company promises to test each new model for dangerous abilities and to add stronger safeguards, or slow down, as its models become more capable.

  1. 1

    Anthropic

    A developer's published framework tying the continued training and deployment of increasingly capable models to evaluations for dangerous capabilities and matching safeguards — in effect a set of if-then commitmentsif-then commitment, n.: A promise by an AI developer of the form: if our model can do a particular dangerous thing, then we will have these protections in place — or we will stop.. The original is Anthropic's Responsible Scaling Policy (September 2023), which defined AI Safety LevelsAI Safety Level, n.: Anthropic's numbered scale, modeled on laboratory biosafety levels, for matching how dangerous an AI model could be with how strong the protections around it must be. modeled on biosafety levels and committed the company not to train or deploy models whose risks outran its protections. Version 2.0 (October 2024) added capability thresholdscapability threshold, n.: A level of dangerous ability — such as meaningfully helping someone make a biological weapon — that, once an AI model reaches it, obliges its developer to add stronger safeguards or hold the model back. paired with required safeguard standards; under it Anthropic activated ASL-3 protections for Claude Opus 4 in May 2025. Version 3.0 (February 24, 2026) restructured the policy: it keeps ASL-3 protections as a unilateral commitment, recasts stronger measures as recommendations for the whole industry, adds a non-binding Frontier Safety Roadmap and Risk Reports every three to six months, and removes language implying that Anthropic would pause if risks became unacceptable. Anthropic presented the rewrite as making risk governance 'proportional, iterative, and exportable'; critics read it as a weakening. Revisions through July 2026 (versions 3.1–3.4) clarified that Anthropic may still pause development, let its Long-Term Benefit Trust request external review of Risk Reports, and adjusted capability thresholds.

    Under its responsible scaling policy, the lab had to show the ASL-3 safeguards were working before launch.

  2. 2

    Generic

    By extension, any frontier developer's framework of this kind, such as OpenAI's Preparedness FrameworkPreparedness Framework, prop. n.: OpenAI's published process for testing its AI models for catastrophic risks — such as help with biological weapons or cyberattacks — and deciding what safeguards are needed before release. and Google's Frontier Safety FrameworkFrontier Safety Framework, prop. n.: Google DeepMind's published plan for spotting when its AI models approach dangerous abilities and applying stronger security and safeguards before that happens.. After the 2024 Frontier AI Safety CommitmentsFrontier AI Safety Commitments, prop. n.: Promises made in 2024 by major AI companies to publish safety plans for their most advanced models, including the danger points at which they would stop if the risks could not be controlled., more than a dozen developers published one, and laws such as California's SB 53SB 53, prop. n.: A 2025 California law requiring developers of the most powerful AI models to publish their safety plans, report serious safety incidents to the state and protect employees who raise alarms. and New York's RAISE ActRAISE Act, prop. n.: A New York law requiring the largest AI developers to publish how they manage catastrophic risks and to report serious safety incidents quickly to a new state office. It takes effect in 2027. now require large developers to publish 'frontier AI frameworks' with similar content.

Origin

Anthropic published the first RSP on September 19, 2023; a week later METR (then ARC Evals), which had helped develop the idea, published a general case for 'responsible scaling policies'. The UK government's request that leading developers publish safety policies before the November 2023 Bletchley summit spread the practice.

Usage note

Critics argue that self-written, self-enforced policies can be loosened once they start to bind, citing the 2026 revision; defenders see them as testbeds for future regulation. Responsible scaling is often contrasted with calls for an AI pauseAI pause, n.: A proposed halt, temporary or indefinite, on building AI systems more powerful than today's, to give safety research and regulation time to catch up..

Related

See also
  • AI Safety LevelAI Safety Level, n.: Anthropic's numbered scale, modeled on laboratory biosafety levels, for matching how dangerous an AI model could be with how strong the protections around it must be.
  • if-then commitmentif-then commitment, n.: A promise by an AI developer of the form: if our model can do a particular dangerous thing, then we will have these protections in place — or we will stop.
  • capability thresholdcapability threshold, n.: A level of dangerous ability — such as meaningfully helping someone make a biological weapon — that, once an AI model reaches it, obliges its developer to add stronger safeguards or hold the model back.
  • Preparedness FrameworkPreparedness Framework, prop. n.: OpenAI's published process for testing its AI models for catastrophic risks — such as help with biological weapons or cyberattacks — and deciding what safeguards are needed before release.
  • Frontier Safety FrameworkFrontier Safety Framework, prop. n.: Google DeepMind's published plan for spotting when its AI models approach dangerous abilities and applying stronger security and safeguards before that happens.
  • Frontier AI Safety CommitmentsFrontier AI Safety Commitments, prop. n.: Promises made in 2024 by major AI companies to publish safety plans for their most advanced models, including the danger points at which they would stop if the risks could not be controlled.
  • SB 53SB 53, prop. n.: A 2025 California law requiring developers of the most powerful AI models to publish their safety plans, report serious safety incidents to the state and protect employees who raise alarms.
Contrast with
  • AI pauseAI pause, n.: A proposed halt, temporary or indefinite, on building AI systems more powerful than today's, to give safety research and regulation time to catch up.

Sources

  1. Anthropic (2023). Anthropic's Responsible Scaling Policy, Version 1.0.
  2. Anthropic. Responsible Scaling Policy updates (version history). anthropic.com/rsp-updates
  3. Centre for the Governance of AI (2026). Anthropic's RSP v3.0: How it Works, What's Changed, and Some Reflections. governance.ai/analysis/anthropics-rsp-v3-0-how-it-works-whats-changed-and-some-reflections

Mentioned in

AI governanceAI governance, n.: The rules, institutions and habits that shape how AI is built and used — from laws and treaties to technical standards and a company's own review processes., AnthropicAnthropic, prop. n.: An American AI company that makes the Claude chatbot and models. It was founded in 2021 by former OpenAI staff and puts particular emphasis on AI safety research., ClaudeClaude, prop. n.: A family of AI assistants and language models made by the company Anthropic since 2023, widely used for writing, analysis and computer programming., Dario AmodeiDario Amodei, prop. n.: An American AI researcher who co-founded Anthropic and is its chief executive, known for early work on scaling and AI safety and for forecasting very rapid AI progress., frontier modelfrontier model, n.: One of the most capable AI models in existence at a given time, built by a handful of leading labs at great cost, and the main target of AI safety rules., loss of controlloss of control, n.: A situation in which people can no longer oversee, correct or switch off an AI system, and have no reliable way of getting that power back., AI whistleblower protectionsAI whistleblower protections, n.: Legal and company safeguards that protect employees of AI companies who report safety dangers or legal violations — to regulators or inside the company — from being fired, sued or otherwise punished., automated AI researchautomated AI research, n.: Using AI systems to do the work of AI researchers, such as running experiments and writing code for new models, so that AI helps build its own successors., capability thresholdcapability threshold, n.: A level of dangerous ability — such as meaningfully helping someone make a biological weapon — that, once an AI model reaches it, obliges its developer to add stronger safeguards or hold the model back., catastrophic riskcatastrophic risk, n.: The risk that AI could cause disaster on a massive scale — many deaths or enormous damage — through weapons, cyberattacks or systems escaping human control. Several laws now define it precisely., dangerous capability evaluationdangerous capability evaluation, n.: A test of whether an AI system can do things that could cause severe harm in the wrong hands or on its own — such as helping make a bioweapon, hacking computers or copying itself., Frontier AI Safety CommitmentsFrontier AI Safety Commitments, prop. n.: Promises made in 2024 by major AI companies to publish safety plans for their most advanced models, including the danger points at which they would stop if the risks could not be controlled., frontier labfrontier lab, n.: One of the handful of companies building the most advanced AI models of the moment, spending billions of dollars on computing power to train them., Jared KaplanJared Kaplan, prop. n.: An American physicist who showed that language models improve predictably as they grow and train on more data, and who co-founded the AI company Anthropic., METRMETR, prop. n.: A nonprofit research group that checks new AI systems for dangerous skills, best known for tracking how long the tasks are that AI agents can now finish on their own., model weight securitymodel weight security, n.: Protecting the files that make up a trained AI model — its learned numbers — from being stolen or leaked, because whoever holds them can run, copy or alter the model., policypolicy, n.: An AI agent's strategy: the rule it follows to pick an action in each situation. Training in reinforcement learning means improving this rule until it earns high rewards., recursive self-improvementrecursive self-improvement, n.: An AI improving its own design or training, so that each improved version is better at making the next improvement. It is the engine usually imagined behind an intelligence explosion., safety casesafety case, n.: A written, evidence-backed argument that a particular AI system is safe enough for a particular use, putting the burden of proof on the developer, as with nuclear plants or aircraft., sandbaggingsandbagging, n.: When an AI deliberately performs worse than it can — for example on a safety test — so that it seems less capable or less dangerous than it really is., SB 1047SB 1047, prop. n.: A 2024 California bill that would have required makers of the biggest AI models to adopt safety plans and take care to prevent catastrophes. It passed the legislature, but the governor vetoed it., SB 53SB 53, prop. n.: A 2025 California law requiring developers of the most powerful AI models to publish their safety plans, report serious safety incidents to the state and protect employees who raise alarms., system cardsystem card, n.: A public report released with an AI model that explains what it can do, how it was tested for safety, what risks were found and what protections were put in place., AI Safety LevelAI Safety Level, n.: Anthropic's numbered scale, modeled on laboratory biosafety levels, for matching how dangerous an AI model could be with how strong the protections around it must be.