structured access
In plain wordsLetting people use an AI system through a controlled channel, such as a website or programming interface, instead of handing over the model itself, so that the developer can limit misuse.
An approach to releasing AI systems in which the developer offers mediated access — through a web interface or APImodel API, n.: A web service through which programs send requests to an AI model run by a company and receive its answers, usually paying per use instead of running the model themselves. — instead of distributing the model's weightsweights, n.: The learned numbers inside a neural network that set how strongly each part influences the next. The finished set of weights is, in effect, the trained model itself., keeping control over how, by whom and for what the system is used. Controls can include usage policies, rate limits, identity verification and tiered access for vetted users, abuse monitoring, input and output classifierssafety classifier, n.: A separate AI model that checks messages going into or coming out of a chatbot and flags or blocks those that break safety rules, like a security guard at the door., and the ability to update or withdraw the model. For researchers, structured access can be extended to deeper interfaces — fine-tuning, activations or gradients — without releasing weights. It occupies the middle of a spectrum between closed and open-weightopen-weight model, n.: An AI model whose trained internal numbers are published for anyone to download and run on their own computers, even if the data and code used to make it stay private. release; supporters see it as a way to share benefits while keeping safeguards enforceable, while critics argue it concentrates power and limits independent scrutiny.
Researchers got structured access to activations through a hosted interface, never the weights themselves.
Origin
Introduced by Toby Shevlane in 'Structured access: an emerging paradigm for safe AI deployment' (2022), a chapter for The Oxford Handbook of AI Governance, which took OpenAI's API-only release of GPT-3 in 2020 as a paradigm case.
Usage note
Structured access concerns the channel; staged deploymentstaged deployment, n.: Releasing a new AI system step by step — first to testers, then to limited groups, then widely — so that problems can be spotted and fixed before everyone has it. concerns the timing. It depends on 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.: once weights leak, access can no longer be structured.
Related
- staged deploymentstaged deployment, n.: Releasing a new AI system step by step — first to testers, then to limited groups, then widely — so that problems can be spotted and fixed before everyone has it.
- open-weight modelopen-weight model, n.: An AI model whose trained internal numbers are published for anyone to download and run on their own computers, even if the data and code used to make it stay private.
- model APImodel API, n.: A web service through which programs send requests to an AI model run by a company and receive its answers, usually paying per use instead of running the model themselves.
- 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.
- misusemisuse, n.: People deliberately using AI to cause harm — for scams, harassment, propaganda, hacking or weapons — as opposed to harm from AI making mistakes or pursuing goals of its own.
- know-your-customer for computeknow-your-customer for compute, n.: Rules that would make cloud companies check who their customers are before renting them enough computing power to train a large AI model, much as banks must verify who opens an account.
- open-weight modelopen-weight model, n.: An AI model whose trained internal numbers are published for anyone to download and run on their own computers, even if the data and code used to make it stay private.
Sources
- Shevlane, T. (2022). Structured access: an emerging paradigm for safe AI deployment. In The Oxford Handbook of AI Governance. Oxford University Press.
Mentioned in
misusemisuse, n.: People deliberately using AI to cause harm — for scams, harassment, propaganda, hacking or weapons — as opposed to harm from AI making mistakes or pursuing goals of its own., biosecuritybiosecurity, n.: Protecting people from the deliberate or careless spread of dangerous germs and toxins; in AI, making sure AI tools do not make it easier to create or release biological weapons., CBRN riskCBRN risk, n.: The danger that AI could help someone make or use chemical, biological, radiological or nuclear weapons — for example, by explaining hard technical steps an attacker could not work out alone., defense in depthdefense in depth, n.: Protecting against AI risks with several independent layers of safeguards, so that if one fails — say, the model's own training — another, such as a filter or human review, still catches the problem., dual-usedual-use, adj.: Describes technology or knowledge that can be used both to help and to harm: the same AI that finds software bugs for defenders can find them for attackers., 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., staged deploymentstaged deployment, n.: Releasing a new AI system step by step — first to testers, then to limited groups, then widely — so that problems can be spotted and fixed before everyone has it., usage policyusage policy, n.: A company's published rules on what people may and may not use its AI for — for example, no help building weapons, no scams and no sexual content involving children., weight exfiltrationweight exfiltration, n.: Copying an AI model's core files out of the place they are supposed to stay — by hackers, by an insider or, in a scenario researchers worry about, by the AI itself.