Anthropic abre em beta a verificação para ciências da vida: salvaguardas de biologia mais flexíveis no Opus 5, Sonnet 5 e Mythos 5.1 para equipas verificadas
- O que aconteceu
- A Anthropic lançou em beta o Life Sciences Verification Program a 17 de setembro de 2026, que flexibiliza as salvaguardas de biologia para equipas verificadas na API e nos planos Enterprise e Team.
- Porque é importante
- O Standard Use abre o Mythos 5.1, o Opus 5 e o Sonnet 5 (e agora o Opus 5.5) ao trabalho corrente de ciências da vida que os classificadores bloqueavam. O High-risk Use vale para o Opus 5 e o Sonnet 5; o Mythos continua limitado a um pequeno conjunto de entidades avaliadas.
- O que fazer
- As equipas de ciências da vida com um plano empresarial da própria Anthropic devem candidatar-se já. Os planos individuais, as plataformas de terceiros e as organizações com BAA ativado ainda não são elegíveis.
Our verdict: if your team does drug discovery, research biology or clinical work on Claude through the API, Claude for Enterprise or a Team plan, apply now. Anthropic's Life Sciences Verification Program (LSVP) went into beta on 17 September 2026. It launched as the first public route to loosened biology safeguards on Claude Opus 5, Claude Sonnet 5 and Claude Mythos 5.1, and Anthropic's program page now lists Claude Opus 5.5 too. No directory rating moves: Opus 5 and Sonnet 5 stay recommended, Opus 5.5 and Mythos 5.1 stay conditional.
What happened
Anthropic opened applications for LSVP, initially for teams and institutions. Organisations that pass a review of research credentials, security standards and ethical research oversight get the biology safeguards that block legitimate requests relaxed. Anthropic says it has already onboarded dozens of organisations through early access and expects to enroll hundreds in the first week.
There are two grants:
| Standard Use | High-risk Use | |
|---|---|---|
| Scope | Most biology research: drug discovery, basic science, R&D, supply chain, clinical development, quality assurance, regulatory affairs and investing | Add-on for work that stays blocked under Standard Use; removes all life-sciences safeguards |
| Granted to | The whole team | A single research project |
| Renewal | Once a year | Every six months |
| Models at launch (17 Sep) | Mythos 5.1, Opus 5, Sonnet 5, plus future models as they launch | Opus 5 and Sonnet 5. On Mythos, limited to a small set of entities with additional vetting |
Why it matters
Before LSVP, the biology safeguards blocked some legitimate research requests, and relief was limited to an early-access group. LSVP swaps real-time blocking for offline monitoring of your usage against the use cases you declare. That makes Opus 5 and Sonnet 5 (and, per the program page, Opus 5.5) usable for day-to-day life-sciences work, and opens High-risk Use on Opus 5 and Sonnet 5 for specific projects.
Mythos 5.1 is the exception. Its directory entry already names Life Sciences Verification as one of its two access routes, and that gate is why it is rated conditional. The gate now has a public application form, but for the highest-risk work Mythos stays behind further vetting while Anthropic works with the US government on broader access.
What changes for you
- Eligibility. The first-party console for API usage, Claude for Enterprise and Team plans. Individual plans are not supported yet, LSVP is not yet on third-party platforms, and BAA-enabled organisations are excluded during the beta. Anthropic suggests teams handling PHI use a separate non-BAA organisation.
- Monitoring. Flagged activity is retained for 30 days for review, kept compartmentalised and excluded from model training.
- Other safeguards. Cyber classifiers and every other safeguard stay on.
- Model choice. The announcement says Standard Use extends to future models as they launch, and Anthropic's program page now lists Claude Opus 5.5 and Sonnet 5 alongside Mythos 5.1. It does not say whether High-risk grants cover Opus 5.5, so confirm that with your account team before you move High-risk work off Opus 5.
FAQ
Can I apply on a Pro or Max plan? Not yet. Anthropic says it is working to expand access to individual plans.
Does a High-risk grant cover my whole team? No. It applies to a single research project and must be renewed every six months.
Can I use LSVP through a cloud marketplace? Not yet. It is not available on third-party platforms during the beta.
O que fazer
- 1 Se é uma equipa de ciências da vida na Claude API, no plano Enterprise ou no plano Team, candidate-se em claude.com/form/life-sciences-verification-program.
- 2 Verifique primeiro a sua configuração face às exclusões: os planos individuais, as plataformas de terceiros e as organizações com BAA ativado não são elegíveis durante o beta. As equipas que tratam PHI devem candidatar-se a partir de uma organização separada, sem BAA.
- 3 Limite qualquer pedido de High-risk Use a um único projeto de investigação e conte com a renovação de seis em seis meses.
- 4 Conte com a retenção durante 30 dias da atividade sinalizada e com a revisão offline da sua utilização face aos casos de utilização que declarar.
Ferramentas e modelos afetados
Nunca mais precisas de te pôr a par
O resumo semanal: apenas mudanças de veredicto e ações urgentes. Sem enchimento.