BRICS Commits to Governing AI — and to Widening Access to It

Eighteen BRICS summits in, and this is the first one where artificial intelligence sat in the declaration rather than in the side events.

The New Delhi Declaration 2026, adopted on Saturday at the 18th BRICS Summit, commits the grouping to implementing its statement on Global Governance of Artificial Intelligence. Short sentence, heavy load. The bloc has moved past calling AI an opportunity and started calling it something that needs governing — collectively, and on terms the members set themselves.

What the declaration actually says

Leaders called for wider access to AI resources and stronger international cooperation, with the stated aim of keeping the technology safe, secure, inclusive and reliable while mitigating its risks. The phrasing matters. "Wider access" is the part that separates this from the governance conversations happening in Washington and Brussels.

The technology agenda in the declaration runs well past AI on its own:

• Artificial intelligence and its global governance
• Cybersecurity and threats arising from the use of ICTs
• Digital public infrastructure
• Telecom networks and submarine cables
• Quantum technologies
• Startups and advanced manufacturing

The declaration also names the harms directly — malicious software, data security failures, cybercrime, misinformation, hate speech, disinformation and deepfakes. Deepfakes turning up in a leaders' declaration is a political choice more than a technical one.

The access question is the real argument

Most AI governance frameworks written in the last three years have been written by countries that already have the compute, the models and the capital. Those frameworks tend to optimise for restraint, because restraint is cheap when you are already at the frontier.

A bloc where most members are not at the frontier has a different problem. For them the risk is not only that AI goes wrong. It is that AI goes right somewhere else, and they end up renting it.

That is why "wider access to AI resources" sits in the same sentence as "safe, secure, inclusive and reliable." The declaration is trying to hold both, and the tension between them is the interesting part. You cannot widen access to compute and models and simultaneously promise tight control over how they get used. Every government writing AI policy right now is negotiating that trade-off. BRICS just did it in public, with twenty-odd signatures on the page.

What it means if you run AI in an enterprise

Declarations aren't regulations, and nothing here changes what your compliance team does on Monday. A few things are still worth tracking.

Sovereign AI stops being a slide. When a leaders' declaration ties AI to digital public infrastructure, submarine cables and advanced manufacturing in one paragraph, it is describing a stack that governments intend to own parts of. Procurement rules follow that thinking eventually.

Governance language converges before rules do. The vocabulary in this declaration — safe, secure, inclusive, reliable — is close enough to what other frameworks use that a company building to one standard is unlikely to be badly wrong under another. The divergence will show up in data residency and access requirements, not in principles.

The risks named are the risks that get resourced. Deepfakes and disinformation appearing in the text means enforcement attention and, eventually, disclosure obligations. If you are deploying generative systems that produce synthetic media or synthetic voices, that is the line to watch.

India chaired this summit, and the declaration reflects a position India has been consistent about: AI capability should be built and shared broadly, not concentrated. Whether the bloc delivers on that is a separate question from whether it has committed to it. The commitment is now on paper.

Source: BRICS calls for wider AI access and global cooperation on safe, secure technology