Acceptable Use Policy

Corollary provides access to protein structure prediction, de novo protein and binder design, genome and protein language models, guide RNA design and general-purpose compute. These are legitimate research tools with a real dual-use profile, and the following uses are prohibited outright:

• Designing, modifying, enhancing or assisting the acquisition of any biological, chemical, radiological or nuclear agent intended or reasonably likely to cause harm to humans, animals, plants or the environment.
• Increasing the transmissibility, virulence, environmental stability, host range or immune escape of a pathogen, or reconstructing an extinct or controlled pathogen.
• Designing toxins, or proteins intended to disrupt an essential biological process in a target organism for a harmful purpose.
• Conferring or evading resistance to therapeutics, vaccines, or medical countermeasures for a harmful purpose.
• Designing or optimising sequences with the purpose or effect of evading biosecurity screening, nucleic acid synthesis screening, or export-control classification.
• Work on agents listed under the Australia Group control lists, the EU dual-use regulation (Regulation (EU) 2021/821) Annex I, or an equivalent national select-agent regime, unless you hold the authorisations that regime requires and can produce them on request.
• Gain-of-function work of concern, or any work your institutional biosafety committee has not approved where approval is required.
• Human germline modification, or the design of reagents intended for it.

If your work is legitimately dual-use — much of pathogen biology and countermeasure development is — it belongs under your institution's biosafety and dual-use oversight, not on a public demonstration platform. Use the appropriate controlled environment instead.

Corollary is not a medical device and has been assessed by no regulator. You must not:

• Use output to diagnose, treat, prevent or mitigate disease in any identifiable person or animal.
• Use output to make or support a clinical, prescribing, dosing, triage or eligibility decision.
• Present output to a patient, or to a clinician for use in care, as if it were validated.
• Submit identifiable patient data, clinical records, or human subject data of any kind.
• Include output in a regulatory submission without independent validation you are responsible for and can defend.

The models and datasets reachable through Corollary belong to third parties and carry their own terms. You must not:

• Use a model or dataset outside the scope its licence permits — in particular, using a non-commercial or academic-only asset for commercial work.
• Use Corollary as a route around a licence condition you could not satisfy directly, including access gating, registration or a click-through agreement on the upstream repository.
• Redistribute model weights, gated artefacts or licensed datasets obtained through the platform.
• Submit content you have no right to submit, including third-party confidential information, or data whose terms forbid transfer to a third-party service.
• Omit the attribution or citation an upstream licence requires when you publish work derived from it.

Corollary itself is a non-commercial demonstration and must not be used for commercial purposes regardless of what any individual model's licence would allow.

You must not:

• Probe, scan, penetration-test or attempt to defeat any security or access control without our written permission, other than good-faith research reported to us in confidence.
• Circumvent credit limits, rate limits or the anonymous allowance, including by creating multiple accounts or rotating network origins for that purpose.
• Scrape the catalogue, the datasets or the site in bulk, or use automated means to place load disproportionate to normal interactive use.
• Use the code sandbox to mine cryptocurrency, host a service, attack a third party, relay traffic, or do anything unrelated to the scientific task at hand.
• Attempt to extract system prompts, credentials, other users' data, or the contents of our infrastructure.
• Introduce malware, or upload content designed to damage or interfere with any system.

You must not use Corollary to break any applicable law, to infringe anyone's rights, to harass or target an individual, to generate material that sexualises children, or to produce disinformation presented as scientific fact.

You must not present AI-generated output as experimentally verified, or as the work of a person or institution that did not produce it. Fabricating results is not merely a policy breach here; in a research context it is misconduct.

Report misuse, or anything on the platform that concerns you, to abuse@corollary-labs.com. Report security vulnerabilities to security@corollary-labs.com.

We may investigate suspected breaches and may suspend or terminate access immediately, without notice and without liability, where we reasonably believe this policy has been breached or where continued access presents a safety, security or legal risk. Where the law requires it, or where there is a credible risk of serious harm, we will report to the relevant authorities.

We monitor for abuse patterns at the level of usage and system behaviour. We do not routinely read the content of your conversations, and we access it only where necessary to investigate a specific report, to fix a specific fault, or where we are legally compelled.

We may update this policy as the platform's capabilities change. Material changes will be posted here with an updated date. Continued use after a change constitutes acceptance.