Claude for Life Sciences: The Verification Program and Its Two Grants

The catch is that this isn't a simple access request. Your grant is tied to the use cases you state in your application, Anthropic monitors for drift from them, your admin gets flagged when it happens, and 30-day data retention is required. That makes it a compliance decision as much as a research one.

Why the general Claude models refuse some biology work

Anthropic says its generally available models, and the announcement names the Fable models, block some life science tasks. Drug discovery, research biology, clinical development and manufacturing are the examples it gives. The programme is meant to let verified teams do that work, from academic labs and startups to pharma companies.

The reason is that in biology it's often impossible to tell valid work from harm. Studying a pathogen to make a vaccine looks a lot like studying it for the wrong reasons. Anthropic says it designed the programme around three threat models:

• Access compromise, such as malware or an account takeover.
• Insider threats.
• Agent misuse, especially by agent swarms or on long-running tasks.

Who gets verified and how

Before opening applications to the wider life science community, Anthropic says it onboarded dozens of organisations through an early-access programme. New applicants join a waitlist. Verification includes a review of research credentials, security standards and ethical research oversight.

Once a team is verified, it can apply for two kinds of grant, and either can be used through Claude Science, Claude.ai, Claude Code and the API. Two early users are quoted in the announcement. Xaira Therapeutics said pairing trusted access with intelligence is the right way to realise AI's promise in biology, and Edison Scientific said the programme lets it bring Anthropic's most intelligent models to its drug-discovery work.

Standard Use and High-risk Use grants side by side

	Standard Use	High-risk Use
What it's for	Most life science work, as daily workloads	Areas that Standard Use blocks
Scope	Can cover an entire team	One research project, not a full team
Renewal	Once a year	Every six months
Safeguards	Classifiers that are more permissive for science tasks	Removes all safeguards that block life sciences requests
Models	Mythos 5.1, Opus 5 and Sonnet 5, plus future models as they launch	Opus 5 and Sonnet 5 at launch; Mythos limited to a small set of entities with extra vetting
How you get it	After verification	As an add-on to a Standard Use grant

Standard Use is meant to cover basic science, R&D, supply chain and manufacturing, clinical development, quality assurance, regulatory affairs, and investing and diligence. For High-risk grants on Claude Mythos, Anthropic said it was working with the US government to make them more widely available.

Anthropic's own example shows how the two fit together. A researcher with dual-use work would typically hold one Standard Use grant for daily activities and one or more High-risk Use grants for specific projects, such as characterising how one family of viral vectors is recognised by human immune pathways.

What stays blocked under a grant

An LSVP grant only loosens the biology safeguards. All other safeguards, such as the cyber classifiers, stay in place.

Monitoring, admin flags and 30-day retention

The safeguards are built around shared responsibility. Each organisation's access is tied to the use cases it states in its application. Anthropic monitors traffic for usage outside that scope and flags cases to the organisation's admins, who are expected to act within pre-agreed timeframes.

That's why the way you write your use cases matters. Anthropic says they should be high-level, like a job listing, with no sensitive information or intellectual property in them.

The programme also moves from real-time blocking to offline monitoring, so legitimate work gets fewer interruptions. To make that work, Anthropic requires 30 days of data retention on LSVP traffic. It says this data is strictly compartmentalised, can't be used for model training and can't be accessed by its life sciences research teams. Anthropic is also working out how the LSVP can integrate with its Enterprise Frontier Safeguards systems for organisations that qualify.

The solutions page, connectors and Claude Science

Claude for Life Science Teams is Anthropic's solutions page for the sector. Its stated approach to verification is claims traced to their source: when Claude connects to PubMed, bioRxiv and research databases, teams can check before they act. It lists built-in biosafety guardrails and says Claude works with full citations and audit trails.

The use cases it names are bioinformatics analysis with connectors to 10x Genomics and Benchling, research synthesis across PubMed, bioRxiv and proprietary databases, clinical trial protocol drafting with regulatory cross-referencing, regulatory affairs submissions, and legacy code modernisation with Claude Code for GxP-regulated environments. If you're new to how Claude plugs into outside tools, my guide to Claude connectors explains the idea.

The page leads with Claude Science, which it describes as being in public beta. It's an app that runs analyses, searches more than 60 scientific databases and traces every step from data wrangling to publication, and my Claude Science guide covers it in full. A banner on the page also says scientists at academic and nonprofit research institutions can get Claude at no cost to start, under a Claude Team plan for scientists.

Applying to the LSVP, step by step

1. Check whether the general Claude models already do your task. If they do, you don't need a grant.
2. Write a high-level use case, like a job listing, with no confidential detail or intellectual property.
3. Apply to the programme and join the waitlist if asked.
4. Expect a review of your research credentials, security standards and ethical research oversight.
5. If you get a grant, name an admin who can respond to monitoring flags within the agreed timeframe.
6. Diarise renewals: once a year for Standard Use, every six months for each High-risk Use grant.

Should your lab or company apply?

I'd apply if the general models keep stopping your team on legitimate drug discovery, clinical or manufacturing work, because that's exactly the gap the programme was built to close. I wouldn't treat it as a sign-up form, though. Before anyone applies, I'd get your compliance lead to agree the stated use cases, sign off the 30-day retention and decide who answers the admin flags. Get those three right and the rest is paperwork.

Frequently asked questions

What is the Life Sciences Verification Program?

It's Anthropic's programme that gives verified life science professionals Mythos, Opus and Sonnet models with safeguards that are more permissive for biology work. As of November 2026 it's in beta, for teams and institutions first.

What's the difference between Standard Use and High-risk Use?

Standard Use covers most life science work for a whole team and renews once a year. High-risk Use is an add-on for one research project, removes the life science blocks and renews every six months.

Does Anthropic keep data from the LSVP?

Yes. Anthropic requires 30 days of data retention on LSVP traffic. It says the data is compartmentalised, can't be used for model training and can't be accessed by its life sciences research teams.

Are cyber safeguards relaxed under an LSVP grant?

No. All other safeguards, such as the cyber classifiers, stay in place under LSVP grants.

Can individuals apply to the LSVP?

Not yet. It launched for teams and institutions, with individual Pro and Max plans to follow over time.