SEOUL, May 21 (Reuters) - Sixteen companies involved in

AI including Alphabet's Google, Meta,

Microsoft ( MSFT ) and OpenAI, as well as companies from China,

South Korea and the United Arab Emirates have committed to safe

development of the technology.

The announcement unveiled in a UK government statement on

Tuesday came as South Korea and Britain host a global AI summit

in Seoul at a time when the breakneck pace of AI innovation

leaves governments scrambling to keep up.

The agreement is a step up from the number of commitments at

the first global AI summit held six months ago, the statement

said.

Zhipu.ai, backed by Chinese tech giants Alibaba ( BABA )

, Tencent ( TCTZF ), Meituan ( MPNGF ) and Xiaomi ( XIACF )

, as well as UAE's Technology Innovation Institute were

among the 16 companies pledging to publish safety frameworks on

how they will measure risks of frontier AI models.

The firms, also including Amazon ( AMZN ), IBM ( IBM ) and

Samsung Electronics ( SSNLF ), voluntarily committed to not

develop or deploy AI models if the risks cannot be sufficiently

mitigated, and to ensure governance and transparency on

approaches to AI safety, the statement said.

"It's vital to get international agreement on the 'red

lines' where AI development would become unacceptably dangerous

to public safety," said Beth Barnes, founder at METR, a

non-profit for AI model safety.

The artificial intelligence (AI) summit in Seoul this week

aims to build on a broad agreement at the first summit held in

the United Kingdom to better address a wider array of risks.

At the November summit, Tesla's Elon Musk and OpenAI CEO Sam

Altman mingled with some of their fiercest critics, while China

co-signed the "Bletchley Declaration" on collectively managing

AI risks alongside the United States and others.

British Prime Minister Rishi Sunak and South Korean

President Yoon Suk Yeol will oversee a virtual summit later on

Tuesday, followed by a ministerial session on Wednesday.

This week's summit will address "building... on the

commitment from the companies, also looking at how the (AI

safety) institutes will work together," Britain's Technology

Secretary Michelle Donelan told Reuters on Tuesday.

Since November, discussion on AI regulation has shifted from

longer-term doomsday scenarios to "practical concerns" such as

how to use AI in areas like medicine or finance, said Aidan

Gomez, co-founder of large language model firm Cohere.

Industry participants wanted AI regulation that will give

clarity and security on where the companies should invest, while

avoiding entrenching big tech, Gomez said.

With countries such as the UK and U.S. establishing

state-backed AI Safety Institutes for evaluating AI models and

others expected to follow suit, AI firms are also concerned

about the interoperability between jurisdictions, analysts said.

Representatives of the Group of Seven (G7) major democracies

are expected to take part in the virtual summit, while

Singapore and Australia were also invited, a South Korean

presidential official said.

China will not participate in the virtual summit but is

expected to attend Wednesday's in-person ministerial session,

the official said.

South Korea's foreign ministry said Musk, former CEO of

Google Eric Schmidt, Samsung Electronics' ( SSNLF ) Chairman Jay Y. Lee

and other AI industry leaders will participate in the summit.

(Reporting by Joyce Lee; Editing by Ed Davies and Christian

Schmollnger)