OpenAI Collaborates With Anthropic and Google DeepMind on AI Safety Frameworks

Key facts

Artificial intelligence developer OpenAI is partnering with corporate rivals Anthropic and Alphabet's Google DeepMind on safety initiatives, according to comments made by OpenAI's global policy chief, Chris Lehane, and reported by Bloomberg News.

While competitive pressures among the trio have intensified rapidly across enterprise software, developer tools, and consumer applications, the companies are seeking common ground on technical safety mechanisms. According to reporting published by Bloomberg News and distributed by Reuters, Lehane outlined that OpenAI is actively working alongside Anthropic and Google DeepMind to address shared safety concerns inherent in building high-capability artificial intelligence.

The announcement reflects a growing practice within the high-technology sector where commercial competitors establish pre-competitive working relationships on safety, security, and interoperability standards while continuing to compete aggressively on product features, model performance, and commercialization. Pre-competitive alignment allows entities to share technical methodologies for measuring model risks—such as accidental misalignment, misuse in cyber operations, or the generation of dangerous biological materials—without forfeiting proprietary intellectual property related to core model architectures or training parameters.

Although specific operational mechanics of the current collaboration were not fully detailed in the initial report, Lehane's statements confirm that dialogue and shared work streams are actively taking place among the policy and technical safety teams at all three organizations.

Why it matters

The establishment of collaborative safety channels between OpenAI, Anthropic, and Google DeepMind carries significant implications for technology governance, enterprise procurement, and global regulatory frameworks. Together, these three organizations represent the overwhelming majority of market share and technical leadership in frontier generative AI deployment.

First, standardizing safety evaluations across competing architectures prevents a destructive "race to the bottom." In fast-moving technological races, individual firms face commercial incentives to accelerate release schedules, potentially truncating safety testing periods. By coordinating evaluation metrics and safety standards, the primary market leaders establish a baseline of responsible deployment that reduces competitive pressure to bypass safety validation.

Second, unified safety definitions provide crucial guidance for global lawmakers and policy enforcement bodies. Regulatory organizations—including the United States AI Safety Institute housed within the National Institute of Standards and Technology (NIST), the European Union's AI Office, and the United Kingdom's AI Safety Institute—rely heavily on technical input from frontier developers. Consistent frameworks from the top three developers reduce regulatory fragmentation and simplify compliance expectations for commercial enterprises integrating AI tools into finance, healthcare, and critical infrastructure.

Finally, joint safety efforts enhance public and institutional confidence in generative tools. As businesses and public sector institutions deploy large language models and autonomous agents into sensitive workflows, verifiable safety standards and independent red-teaming protocols become mandatory prerequisites for widespread adoption.

The landscape of artificial intelligence safety has evolved rapidly alongside the escalation of model capabilities over recent years. OpenAI, founded in 2015 as a non-profit research laboratory before creating a profit-capped commercial entity, brought generative AI into mainstream global awareness with the public release of ChatGPT in November 2022. The company's rapid growth and shifting corporate governance structure previously sparked internal debates regarding the balance between rapid commercial deployment and safety research.

Anthropic was established in 2021 by former OpenAI research leaders, including siblings Dario Amodei and Daniela Amodei, who departed OpenAI specifically to build an organization focused primarily on AI alignment, safety research, and system steerability. Anthropic developed the Claude series of language models and pioneered architectural safety techniques such as "Constitutional AI," which trains models to adhere to explicit sets of ethical principles during fine-tuning.

Google DeepMind represents the consolidated AI research powerhouse of Alphabet Inc. Formed through Google's 2014 acquisition of London-based DeepMind and its subsequent April 2023 merger with the Google Brain division, the entity is led by Demis Hassabis. Google DeepMind has consistently emphasized fundamental AI safety, publishable alignment research, and scientific applications of deep learning alongside its commercial Gemini model family.

Industry-wide safety coordination began formalizing in mid-2023. In July 2023, seven leading AI companies—including OpenAI, Anthropic, Google, and Microsoft—signed voluntary commitments at the White House to undergo external safety testing, watermark synthetic content, and publicly report model capabilities and flaws. Shortly thereafter, in July 2023, OpenAI, Anthropic, Google, and Microsoft launched the Frontier Model Forum, an industry body formed to promote safe and responsible development of frontier AI models.

Global governments have simultaneously escalated formal oversight. The United Kingdom hosted the inaugural International AI Safety Summit at Bletchley Park in November 2023, resulting in the Bletchley Declaration signed by 28 nations. The European Union formally enacted its comprehensive AI Act in 2024, imposing tiered safety and reporting obligations on developers of high-impact general-purpose AI models.

Reaction

Following the report of Lehane's comments, tech industry analysts, policy researchers, and government oversight bodies are monitoring how the joint effort will translate into verifiable safety practices.

While neither Anthropic nor Google DeepMind immediately issued separate public statements confirming specific new joint initiatives beyond existing industry forums, regulators generally welcome technical coordination among market leaders. Officials at government safety bodies, including the U.S. and U.K. AI Safety Institutes, have routinely encouraged frontier labs to share red-teaming methodologies, evaluations for extreme risks, and vulnerability disclosures.

Conversely, civil society organizations, academic ethicists, and open-source software advocates often maintain a cautious stance toward self-regulatory agreements among corporate dominant players. Critics frequently emphasize that voluntary lab-led safety coalitions must remain transparent and subject to independent third-party auditing to avoid functioning as mechanisms for regulatory capture or anti-competitive market concentration. Antitrust regulators in both the U.S. and Europe also maintain strict oversight to ensure that legitimate pre-competitive safety coordination does not cross into unauthorized commercial collusion or market partitioning.

What we don't know yet

Despite the confirmation of safety collaboration by OpenAI's global policy chief, several technical and structural details remain unclarified in the available reporting:

The resolution of these open questions will determine whether the joint safety effort represents a meaningful advance in cross-industry standardization or a routine executive dialogue.

What to watch

In assessing the future trajectory of this cross-industry collaboration, key developments to watch include:

This report is based on original reporting published by Bloomberg News and distributed by Reuters, as reported by The Star.