Best AI Evaluation Companies

Vendor profileChecked Nov 8, 2026

Apollo Research

Apollo Research (states it is now a Public Benefit Corporation) · apolloresearch.ai · Last checked:

  • Independent evaluator and research organization
  • Safety evaluations and pre-deployment testing

Company status

Active. States it is now a Public Benefit Corporation and has opened a San Francisco office.

Sources 1, 2, 3, 4 · checked 2026-11-08

Capabilities

  • Pre-deployment safety evaluations of frontier AI systems, focused on scheming, strategic deception, sabotage, sandbagging, and evaluation awareness
  • Assessment sections published in developers' system cards (OpenAI, Anthropic, and Meta models listed on its science page)
  • Technical AI governance support for governments and international organizations
  • Watcher: a separate runtime monitoring and control product for AI agents

Sources 1, 2, 3, 4 · checked 2026-11-08

Who it tends to fit

  • Frontier model developers wanting third-party scheming and deception evaluations before release
  • Governments and international bodies working on technical AI governance

Sources 1, 2, 3, 4 · checked 2026-11-08

Pricing

Contact vendor (partnership inquiry form)

Sources 1, 2, 3, 4 · checked 2026-11-08

Published certifications

None found / not verified. This is not a finding that Apollo Research lacks certifications; ask the vendor for documentation.

No security certification found on the official site on 2026-11-08. Ask for security documentation.

Listed as stated by the vendor. Scope can differ by product or subsidiary; request current reports.

Sources 1, 2, 3, 4 · checked 2026-11-08

Limitations to check

  • No public pricing or published engagement terms
  • Scope is focused on scheming and deception risks rather than general capability or quality testing
  • No security certifications published

Sources 1, 2, 3, 4 · checked 2026-11-08

Sources

  1. Science page: 'We also develop and run pre-deployment evaluations of frontier AI systems'; lists system card contributions for OpenAI, Anthropic, and Meta models covering deception, scheming, sabotage, sandbagging, and evaluation awarenesswww.apolloresearch.ai/scienceChecked 2026-11-08
  2. Homepage lists pre-deployment evals, governance work, the Watcher product, and a partnership inquiry routewww.apolloresearch.ai/Checked 2026-11-08
  3. About page: 'We are now a Public Benefit Corporation'; opened an office in San Francisco; ran evaluations for all major labs in 2025www.apolloresearch.ai/aboutChecked 2026-11-08
  4. Governance page: supports governments and international organizations in developing technical AI governance regimeswww.apolloresearch.ai/governanceChecked 2026-11-08

Report a correction: [email protected]. Facts are re-checked at least every 90 days.