What do startups want from the AI Safety Summit?

As the UK government prepares to host the AI Safety Summit between November 1 and 2, startups are calling for their interests to be heard as big tech and world leaders gather at Bletchley Park to discuss regulations to govern artificial intelligence.

With a focus on frontier AI, the Summit represents a bid by the UK government to convert the country into a natural hub for efforts to regulate artificial intelligence on a global scale.

Leading academics and executives from AI companies, including Google’s DeepMind, Microsoft and OpenAI will gather to drive targeted action to develop policies conducive to the safe and responsible development of AI.

However, AI startups are concerned that their interests will be sidelined in favour of larger tech players with more resources and capital, despite being at the frontline of product development and innovation.

Why startups matter in the AI safety conversation

AI development has been dominated by the likes of OpenAI and Meta, responsible for the release of ChatGPT and the open-source Large Language Model Llama 2. However, within the UK, AI startups have played an important role in the technology’s roll out and innovation.

According to data from the Department of Science, Innovation and Technology (DSIT), 88% of the AI business population in the UK is made up of small (10-49 employees) or micro (1-9 employees) businesses.

Although they only make up 28% of AI’s economic contribution, they represent the majority of the AI business landscape in the country. Any guardrails that alter how AI develops in the UK will have a larger impact on the AI small business community.

“The startup’s voice is especially important to ensure regulation is protective of and non-burdensome to smaller tech companies and AI entrepreneurs,” stresses Yan Zhang, COO of startup PolyAI.

“It’s not surprising that legislative action in business tends to favour the larger players, but for as ubiquitous as AI in the enterprise is bound to be, it only makes sense to create a consortium of lesser-financed voices to have a seat at the table.”

Other founders not only agree about the importance of having startup voices at the Summit, but note the unique value startups can contribute to fostering inclusive and holistic policies.

“AI Startups are the companies who are defining the future,” stresses Rafie Faruq, CEO and Co-Founder of legal startup Genie AI. “It’s also important to seek viewpoints from startups that are diverse in their talent pool, which is normally the case for startups since they typically focus on their company culture more than large incumbents.”

Faruq believes the government should establish panels, inclusive committees, representation at meetings and conferences, and feedback mechanisms so the startup community can be adequately represented.

As AI continues to evolve, founders say democratising the sector is key to guarantee its innovative health. “Many people believe that AI safety regulations are a cynical ploy towards regulatory capture of the sector by Big Tech,” shares James Clough, CTO and Co-Founder of Robin AI.

“Including earlier stage startups is one way to avoid that occurring, and that perception which may otherwise undermine support for important safety regulation.”

What type of AI regulations do startups want and need?

Although AI startups work across different industries, from the legal sector to biomedicine, there is a concerted desire to have concrete regulations that don’t stifle innovation, as well as policies that grant access to data that can help train AI systems.

“Governments have vast amounts of trainable information in bureaucratic vaults, and granting access to this data would benefit citizens, businesses and startups alike,” notes Zhang.

“The lack of a data union (akin to a customs union) is a significant hurdle preventing collaborative development in AI,” he adds.

Data is a common theme of concern for startup owners. “One thing that could help startups and large businesses alike is clarification on Copyright law,” outlines Faruq.

“AI models learn from public data, and what data is subject to copyright is often a grey area, especially since copyright law evolved out of property law many years ago, and in my view isn’t fit for modern times.”

Are AI regulations the right solution?

Whilst startup founders acknowledge that regulations of some sort are needed, particularly to govern access to data, some are more sceptical about the emphasis the UK is placing on developing them quickly to become a regulatory leader.

“The right way to interpret ‘leader in AI safety regulations’ should not be ‘first’ – but rather, ‘best’,” stresses Clough. “In order to be effective, safety regulation cannot be seen to be Luddite in nature, and so must be focused only on serious risks.”

The danger that could emerge from the Summit is that regulations are put on a pedestal as the ultimate solution and end up restricting the room for innovation within AI.

“Everyone is talking about AI regulation (which is the solution), but what problem is this solving? Depending on the problem, different solutions could be needed, and the solution may not be regulation at all, even though that’s the buzz word that’s being thrown around at the moment,” warns Faruq.

Startup founders emphasise that AI is evolving quickly and regulations are required to an extent. However, the Summit should focus only on the most pressing safety concerns rather than trying to reach a consensus on a broad range of issues.

Beyond the AI Safety Summit

The AI Safety Summit will be a crucial focal point for the industry, but startups want their interests to be considered beyond the conference.

Thus far, the government has established channels for funding and support of AI startups, which are praised by founders. These include the InnovateUK program, which seeks to transform the UK into a global hub for innovation by 2035 by helping businesses to access the expertise and equipment they need to grow.

As the technology landscape continues to evolve through artificial intelligence, startups want to see evidence of the government adapting to these changes.

“Now that the times have changed, the government needs to focus on more critical infrastructure to support AI,” emphasises Faruq, “This includes investment in AI talent on all levels – both in startups and at universities, otherwise the talent is eaten up by large incumbents.”

Others call for greater access to R&D credits, given how scant access to tax credit has been since regulations changed following the Spring Budget announcement.

The government’s discourse, considering the release of the AI whitepaper and Michelle Donelan’s speech at CogX 2023, reveals the government is prioritising regulations in its bid to become a technology hub for artificial intelligence.

If this mission is to become a success, the voices of startups should be included in the conversation not only to democratise the development of the technology but to ensure that all regulations have factored in all the stakeholders they will affect.

As world leaders, academics, and big tech gather at Bletchley Park for the Summit, startups will not only hope to be included in the conversation but to have policy makers understand the importance of only using safety regulations where they are justified. If not, the speed of innovation of benevolent AI could stagnate.

Read more:

• The King’s Speech: what can businesses expect from the Autumn Statement?