OpenAI, Anthropic and Google reportedly explore joint AI standards body

OpenAI, Anthropic and Google are reportedly exploring the creation of an artificial intelligence (AI) standards body, potentially bringing three of the industry's biggest rivals together to establish safeguards around increasingly powerful AI systems.

According to The Straits Times, OpenAI confirmed on 15 September that it was working with Anthropic and Google on the idea, which originated from a proposal by Google DeepMind CEO Demis Hassabis.

The proposed self-regulatory body would reportedly be modelled after the Financial Industry Regulatory Authority (FINRA), which oversees brokers and investment firms in the US, and could independently test the most powerful AI systems before they are released to the public.

Don't miss: Anthropic builds Singapore footprint with hiring push

Hassabis reportedly first called for such an organisation in July, arguing that it would require substantial funding, likely largely from the industry, to attract technical talent and secure the computing resources required for large-scale testing.

MARKETING-INTERACTIVE has reached out for more information. 

The discussions come as leaders across the AI industry increasingly debate whether the rapid development of frontier models needs to be deliberately slowed to give safety measures more time to catch up.

Anthropic CEO Dario Amodei recently called on AI companies to "pace the frontier", warning that rapidly advancing capabilities could outstrip companies' ability to safely understand and control their systems.

Rather than stopping AI development altogether, Amodei proposed a three-stage framework beginning with ongoing access for independent third-party evaluators to assess safety practices, report incidents and scrutinise both finished models and training processes.

The second stage would see frontier AI companies in democratic countries coordinate around common safety standards and limits on unchecked AI progress, before eventually pursuing broader coordination between governments internationally.

OpenAI CEO Sam Altman backed Amodei's call, including his proposal to give independent evaluators employee-like access to AI companies. Altman said OpenAI would make the same commitment, while xAI founder Elon Musk also publicly backed Amodei's position.

Who gets to write the AI rulebook?

However, there is yet to be consensus over how an industry-led standards system should operate, with the prospect of some of the world's largest AI companies collectively shaping the rules drawing criticism from elsewhere in the sector.

Cohere co-founder and CEO Aidan Gomez has argued that while AI needs guardrails, those standards should not be determined by a small group of dominant technology companies.

In a blog post published on 13 September, Gomez said the dispute was not over whether safeguards were necessary, but "who writes them, who gets to participate and whose interests the rules are protecting".

Gomez took particular issue with proposals for leading AI companies to coordinate on shared standards and limits on the pace of development, arguing that allowing competitors to do so under an exemption from competition rules could entrench today's largest players.

He instead called for an evidence-based approach involving governments, technologists, academics, civil society, smaller AI labs and experts from sectors such as finance and critical infrastructure.

His proposed alternative rests on four areas: developing an evidence-based framework for defining AI risks, requiring greater transparency from developers, independently testing systems according to their capabilities and potential harms, and establishing assurance mechanisms independent of the companies being assessed.

Gomez also argued that safeguards should be determined by what an AI system can do and where it is deployed, rather than the size of the company developing it or the computing resources used to build it.

The debate over how AI should be governed comes amid increasingly stark warnings from researchers who have worked inside some of the industry's leading labs.

Former Google DeepMind researcher Bilal Chughtai, who recently resigned from the company after working on artificial general intelligence (AGI) safety and alignment, said he was "extremely concerned" about the current trajectory of the technology.

In a post on X, Chughtai argued that frontier AI capabilities were advancing faster than researchers' understanding of alignment, warning that increasingly capable systems could become difficult to control.

He called for greater coordination between AI companies to slow development to a pace at which emerging risks can be addressed, alongside greater transparency into how AI systems are being developed.

His concerns echo those raised by Jacob Coxon, who recently left Anthropic after previously working at OpenAI and warned about the potential risks of increasingly powerful AI systems, according to The Straits Times.

However, not everyone in the industry agrees that slowing development is the answer. Nvidia CEO Jensen Huang has argued that AI safety and speed do not have to come at the expense of one another, describing safety as an "engineering problem".

Speaking at Dreamforce this week, Huang said companies should thoroughly test their systems and hold back products they cannot demonstrate are safe but rejected calls to slow AI innovation more broadly.

"I think innovation, speed and safe products is a false choice. You could definitely have both at the same time. So run as fast as you can," Huang said.

Related articles: 
Malaysia’s Digital Ministry puts AirAsia's Aireen Omar at helm of AI Malaysia   
Singapore to create registry for AI agents   
Public trust in AI rises as media becomes more sceptical, CARMA finds