AI Safety Rankings 2026: No Lab Gets an A, Regulation Looms

The latest AI Safety Index from the Future of Life Institute, released February 28, 2026, gives no AI lab an A grade. Anthropic tops the AI safety rankings 2026 with a C+, while OpenAI drops to C and Google DeepMind stays at third. Meta improves from D to D+, and SpaceXAI falls to an F alongside DeepSeek and Mistral. The index shows that even the best-performing labs have weak safety practices, and the institute's co-founder says regulation is needed to drive real improvement.

What happened

The AI Safety Index, produced twice a year, asks an expert panel to grade AI firms on how seriously they manage risk from pre-deployment testing to plans for keeping ever-more-powerful systems under control. The grades are unforgiving.

All three top labs have weakened or dropped earlier pledges to halt development if certain red lines came into view, and they softened their resistance to military uses. In February, Anthropic dropped its pledge to never train an AI system unless it could guarantee in advance that its safety measures were adequate, a measure the panelists recommended reversing Source: TIME article. Meta climbed from D to D+ in six months, rising from sixth place to fourth. Max Tegmark, the institute's co-founder, said "It's encouraging to me that a company can improve so much in just six months."

SpaceXAI fell to an F after rebranding from xAI, joining China's DeepSeek and France's Mistral. Tegmark noted that the worst scorers come from three different continents, showing this is a global problem.

Why AI builders should care

These AI safety rankings 2026 reveal that even the most safety-conscious labs (Anthropic at C+) cannot achieve an A-grade safety posture. For builders and product teams, this means you cannot rely on a lab's brand alone to ensure safe deployment. The index's criteria pre-deployment testing, governance, and safety controls apply directly to any team building with or on top of foundation models.

Regulation is also accelerating. Tegmark pointed to the EU AI Act, Chinese rules taking effect later this month, and a more risk-conscious U.S. administration as signs of progress toward global safety standards Source: TIME article. These regulations will likely require verifiable safety commitments from all AI builders, not just the largest labs.

Practical implications

For AI builders, the rankings suggest three practical actions. First, map your product stages to safety milestones aligned with the Frontier AI Safety Commitments and similar governance best practices. Second, adopt transparent, verifiable safety testing protocols rather than trusting vendor claims. Third, prepare for regulatory scrutiny by documenting your risk management practices, model evaluations, and deployment safeguards.

Meta's improvement from D to D+ in six months demonstrates that meaningful progress is possible. Builders who invest in safety governance now may have a competitive advantage as regulations tighten.

Caveats

The AI Safety Index is based on the Future of Life Institute's methodology as of Winter 2025/2026. Exact scoring criteria and grades may vary across releases. Some companies have publicly changed their pledges or constraints, which can affect their positions in subsequent indexes. The rankings reflect the panel's assessment of voluntary commitments, not necessarily legal compliance. Builders should monitor the official index releases for the most current data.

FAQs

What is the AI Safety Index and how is it scored?

The AI Safety Index is a twice-yearly ranking by the Future of Life Institute that grades AI companies on how they manage risk. It evaluates pre-deployment testing, governance of powerful systems, and plans for maintaining safety controls Source: TIME article.

Which companies topped or bottomed the latest AI safety rankings?

Anthropic topped the latest AI safety rankings 2026 with a C+. OpenAI dropped to C, and Google DeepMind ranked third. Meta improved to D+ (fourth), while SpaceXAI (formerly xAI) fell to an F, shared by DeepSeek and Mistral Source: TIME article.

Why did no company receive an A in the AI safety rankings?

No company received an A because even the top labs have significant gaps in safety governance. The index panel found that all three leading labs weakened or dropped earlier pledges to halt development under red-line scenarios and softened their resistance to military uses Source: TIME article.

What factors influence AI safety ratings?

Factors include pre-deployment testing, governance of powerful systems, and plans to keep safety controls in place. The panel also considers whether companies have weakened their safety pledges or softened resistance to military uses Source: TIME article.