Anthropic Proposes Transparency Framework For AI Model Development

The AI company Anthropic proposed a transparency framework on Monday for advanced AI models and companies developing frontier AI systems, intended for application at regional or international levels. The startup outlined safety measures, actionable steps, and minimum standards to enhance AI transparency.

Anthropic explained in an announcement on its website that the development of AI models has progressed more rapidly than the creation of safeguards and agreements by companies, governments, or academia. The tech company urged all stakeholders to accelerate their efforts to ensure the safe development of AI products and offered its transparency framework as a model or reference.

“We need interim steps to ensure that very powerful AI is developed securely, responsibly, and transparently,” states the announcement. “We are therefore proposing a targeted transparency framework, one that could be applied at the federal, state, or international level, and which applies only to the largest AI systems and developers while establishing clear disclosure requirements for safety practices.”

Anthropic’s approach has been simplified to keep it flexible and lightweight. “It should not impede AI innovation, nor should it slow our ability to realize AI’s benefits—including lifesaving drug discovery, swift delivery of public benefits, and critical national security functions,” clarified the company.

The tech company acknowledged that rigid frameworks and standards could quickly become outdated as the technology continues to advance at a rapid pace.

Anthropic suggests that AI transparency requirements should apply only to large frontier model developers, in order to avoid burdening small startups and low-impact developers. The proposed threshold is $100 million in annual revenue or $1 billion in yearly capital expenditures.

Large developers should also create a public Secure Development Framework that includes how they will mitigate risks, including harms caused by misaligned models and the creation of chemical, nuclear, biological, or radiological weapons.

One of the strictest proposals is aimed at protecting whistleblowers. “Explicitly make it a violation of law for a lab to lie about its compliance with its framework,” wrote the Anthropic in the document shared. The company emphasized that transparency standards should include a minimum baseline and remain flexible, with lightweight requirements to help achieve consensus.

Anthropic expects the announcement and the proposed transparency framework to serve as guidelines for governments to adopt into law and promote “responsible practice.”

After releasing its latest AI model, Claude 4, Anthropic included a safety warning—labeling the model Opus 4 on Safety Level 3 due to its potential risks.