OpenAI Calls for Global AI Safety Standards, but Still Releases New Models

OpenAI called for the establishment of international AI safety and security standards as well as increased research for responsible technology development. The call was made in a blog post on Monday, amid concerns that the pace of AI innovation has outpaced existing safety frameworks.

In its statement, OpenAI emphasized that navigating the next stage of AI development requires "alignment research that can keep pace with AI capabilities, so that the systems we and others build remain aligned with human values and under human control." The statement came a week after CEO Sam Altman, along with leaders from Anthropic, Microsoft, and SpaceXAI, called for a slowdown in the development of frontier AI models. Altman and Anthropic CEO Dario Amodei are scheduled to brief the UN Security Council this week.

OpenAI argues that international standards are as important as research in governing the pace of technology. Such standards can help address challenges facing AI providers, such as fragmented safety evaluations and uneven AI development capacity across different parts of the world. The company also emphasized the need to strengthen digital security defenses, advance alignment research, and help the public understand and steer the next stage of AI development.

"This becomes increasingly important as AI systems grow more capable and autonomous, including when they take over more of the research and development of AI itself," OpenAI wrote in the blog post. The company cited institutions such as the Center for AI Standards and Innovation and state laws as potentially helping standardize research efforts. A federal framework for AI and forms of public-private partnership are also seen as potentially accelerating the creation of harmonized standards. Last June, President Donald Trump signed an executive order requiring AI providers to undergo voluntary reviews of their models, but there is still no official binding federal framework.

International efforts to develop evaluation standards, reporting requirements, and incident definitions can help researchers compare evidence, understand emerging capabilities, and respond to risks globally. "AI development and deployment happen in many countries, its adoption is global, and its impacts will likely be felt by everyone in some form at some point," OpenAI added.

Despite calling for a slowdown, OpenAI instead released GPT-6 Sol and DPT-6 Luna on Tuesday. The cheaper models are aimed at advancing professional work. The move raises questions about the consistency between the safety call and the company's commercial strategy.

Amid global calls to regulate the pace of AI research, corporate CIOs have more control over what happens inside their organizations than they realize, according to CJ Combs, senior sales director for data and AI at Columbus Global US, a consulting and IT services firm. "The call is usually framed as a pause for labs until their models are safer," Combs told CIO Dive via email. "The pause that actually matters happens inside the enterprise, and it's a pause to build, not to wait."

Many organizations deploy AI without the right foundations, Combs said — on top of unclean data, immature governance, and without clear ownership. They are then surprised when their AI agents act in ways they cannot explain. Slowing down to fix data, defining who owns oversight when agents cross HR, finance, IT, or security boundaries, and setting value targets are ways for CIOs to control safety and security standards within their organizations, Combs explained.

OpenAI's call for standards is a good signal, but it shows more about how AI providers plan to behave than about how technology leaders should model AI use within their organizations, he said. Providers cannot see an organization's data, approval chains, or which systems interact with agents. "Both can be true. Global standards are worth building, and CIOs still cannot outsource responsibility to those standards," Combs said. "Treat any external standard as a floor and set your own higher."

For Indonesia, this development serves as an important mirror. As a country with rapidly growing AI adoption, Indonesia does not yet have a comprehensive AI regulatory framework. OpenAI's call for international standards could push the government to accelerate the drafting of rules, while also reminding domestic companies that compliance with global standards is not enough — they need to build strong internal governance. Without that, the risks of AI misuse, data leaks, and uncontrolled automated decisions will become increasingly real.

The question now: will OpenAI's call be followed by real action, or will it remain rhetoric amid an increasingly tight commercial race? And for Indonesia, can we craft an AI framework that not only pursues international standards but also protects national interests?