UAE Official Calls For Independent AI Safety Checks Amid Rapid Advances

(MENAFN- Khaleej Times) Countries should stop expecting the global artificial intelligence (AI) race to slow down and instead build independent systems for testing increasingly powerful models, a senior UAE official said on Tuesday.

“Nobody is going to slow down,” Faisal Al Bannai, adviser to the UAE President and secretary general of the Advanced Technology Research Council, said during a conversation with CNN's Becky Anderson at AI Everything Abu Dhabi.“At the end of the day, everyone is pushing ahead at full steam.”

Recommended For You

Al Bannai said the more practical question was how governments and independent bodies test frontier models as they become more capable, particularly when they are placed inside critical systems.

Stay up to date with the latest news. Follow KT on WhatsApp Channels.

He argued that the organisation building an AI model should not also be solely responsible for declaring it safe.

Instead, he called for independent reviews, common benchmarks and safety standards that evolve alongside the technology rather than conventional regulations written once and left unchanged for years.

The level of scrutiny should also depend on where AI is being used.

“It's one thing having an AI write your email,” Al Bannai said.“It's another thing having an AI decide which tool goes in the hospital, what happens in traffic. That can really affect people's lives.”

What does 'sovereign AI' actually mean?

The discussion also went deeper into what Al Bannai believes countries mean when they talk about sovereign AI.

Owning intellectual property alone, he argued, is not enough. A country must understand the underlying technology and retain the ability to modify, test and evolve it.

“If you do not have the control of evolving it, testing it, modifying it, and really driving the destiny of that technology, it's not sovereign,” he said.

That becomes particularly important when AI begins controlling physical systems.

If a country integrates somebody else's model into a robot without understanding what is happening inside it, Al Bannai said, it may be able to perform the required job, but it cannot fully understand its limitations or dangerous edge cases.

“If you don't know what's inside that code, if you're unable to control the weights of that AI, then frankly, you're a hostage to someone else giving you a black box,” he said.

The UAE's approach, he said, is not to choose exclusively between international and locally developed models.

“We're adopting the best-in-class closed models, and we're building our own open-source models,” he said.

He also disclosed that the Technology Innovation Institute is developing a product called AgentGuard , which he described as another layer of“circuit breakers” and filters intended to prevent AI agents from performing actions they should not take.

Al Bannai acknowledged that such safeguards would not capture every possible failure.

“Will it cover all corner cases? No. It's better than nothing, and you keep evolving it from there.”

Smaller models for real problems

That pragmatism is also changing the type of AI being developed in Abu Dhabi.

Al Bannai said TII had initially competed at the frontier of large language models before shifting greater attention towards smaller and more specialised systems.

“Seventy, 80 per cent of what we all need in enterprise is not an extremely large model that can boil the ocean,” he said.

Instead, many companies need smaller models designed around specific tasks in areas such as transport and other industries.

TII's recent model development reflects that direction. Its Falcon-H1 family ranges from compact 500-million-parameter systems to a 34-billion-parameter model, while subsequent Falcon-H1-Tiny research has pushed specialised models down to as few as 90 million parameters for applications where compute, latency and energy use matter.

Al Bannai said the UAE's competitive advantage was therefore not simply an attempt to build a larger general-purpose model than the US or China.

Instead, he pointed to the country's ability to make decisions quickly, allow companies and government entities to test new applications and adapt regulation as technology develops.

“The key advantage of the UAE is having a regulator that is bold enough to understand, to understand the risks, and to move with agility,” he said.

That approach inevitably raises questions about checks and balances when government can simultaneously be an investor, regulator, customer and builder of technology, a point Anderson put directly to him.

Al Bannai acknowledged the tension, but rejected the idea of waiting until governments discover a perfect formula for AI governance.

“The genie is out of the box,” he said.“You will not be able to just say, let's wait. Because nobody will wait.”

The remarks came at AI Everything Abu Dhabi, a two-day AI summit and exhibition taking place at Adnec Centre on October 6 and 7.

ALSO READ

Tech expert explains AI risk, what UAE residents should do amid industry concerns Excessive focus on AI regulation can overshadow its practical benefits: UAE minister AI can accelerate cybersecurity but humans must retain the final decision

MENAFN06102026000049011007ID1111766991