Ada Lovelace Institute

'Open-source AI' can be a loaded term. Many AI models, systems and frameworks are equally labelled as open source when they offer different degrees of access. A weak definition of open-source AI, with no requirement for open training data, could benefit larger companies over independent AI development and innovation. In our new blog post Mike Nolan unpacks the Open Source Initiative's definition of open-source AI, questions its potential shortcomings and reflects on how it might interact with AI regulation, such as the EU AI Act. https://lnkd.in/eu85KjS3

I spoke to the FT about this morning's news about the Government's Open AI deal - hot on the heels of the announcement earlier this month of a new partnership with Google. Our research shows that public concerns about AI systems in the public sector, such as for welfare eligibility or predicting the risk of cancer, are higher when these are produced by private companies rather than government bodies, NGOs or academics. We've repeatedly seen concerns raised about the role and influence of 'big tech' in discussions about public sector AI, from health to policing. Far from easing public concerns, these types of deeper partnerships may, therefore, reinforce the perception that AI in the public sector means handing more power to private companies. People may reasonably wonder whether, rather than taking actions to address the power of big tech, the government is helping entrench it. That's why we made a recommendation (in Licence to Build - link in comments) that the Science, Innovation and Technology Committee to launch an inquiry into the private sector’s involvement in shaping government policy and procurement decisions related to AI in the public sector. https://on.ft.com/46oNOBW

🤩 I’m honoured to have been included in ComputerWeekly.com's UK Tech 50. https://lnkd.in/eVW72_J9  🤔 It’s always nice being recognised on a list like this – but what really struck me was how it broadens out what it might mean to ‘influence’ the tech ecosystem in the UK. 🫂 It’s not just big companies that have the power to shape the trajectory of technology, but also fellow researchers, social entrepreneurs, civil servants, third-sector leaders and campaigners.  🤝 At Ada we’re fortunate to be in a position where we can bring diverse groups together to ensure that AI and data-driven technologies work better for people and society: from government ministers, to front-line workers, to the public. 📣 Convenings like this will be a cornerstone of our new strategy, which will launch soon. 👀 Watch this space.

The European Commission has received the final version of the General-Purpose AI Code of Practice, a voluntary tool designed to help industry comply with the AI Act's rules on general-purpose AI, which will come into effect on 2 August 2025. The final version of the Code of Practice is a mixed success. It represents one of the most robust tools we have so far for describing effective risk mitigation approaches; however, policymakers will need to be much more ambitious in future iterations. Read our full response ⬇️ https://lnkd.in/ewuXRj8v

The UK government is set to create a governance framework for police use of facial recognition technology. Ada's Nuala Polo discusses these plans in the article below. Read our report which calls for a more comprehensive biometrics governance framework that also includes high-risk uses like private sector surveillance and inferential biometrics – not just police use. https://lnkd.in/ePCueeqb

AI assurance allows us to measure, evaluate and communicate the trustworthiness of AI systems. But what are the conditions needed for, and potential impacts of, professionalising the AI assurance industry?   Ada and the Center for Democracy & Technology’s new research found several gaps that the AI assurance field must address to ensure robust and adaptable practices.   Professionalisation of AI assurance could help to develop safe, reliable and effective AI systems that deliver benefits for people and society.   As AI technologies and their applications continue to evolve, a multistakeholder coalition of policymakers, industry, civil society and more is needed to support safe innovation in AI.   It is also crucial that efforts to professionalise the AI assurance industry incorporate evidence from experts on the ground.   Read 'Going pro?' https://lnkd.in/ecJiU_XR

Really pleased our chapter for the British Social Attitudes report has been published today. We found that public attitudes to AI are influenced by political values, particularly in relation to concerns about discrimination and job losses. In particular, applications of AI in the public sector vary significantly by political orientation, perhaps because of the impact these uses have on people’s lives and liberty. Importantly, support for laws and regulations around AI are not as strongly associated with political orientation as concerns. Around 7 in 10 people say they would feel more comfortable with AI technologies if they were governed by laws and regulations, a sentiment held widely across political orientations Tvesha Sippy, Helen Margetts, Oliver Bruff, Jorge Perez, Aditi Dutta. Ada Lovelace Institute The Alan Turing Institute

Across our research with the public, we've repeatedly heard concerns about AI in the public sector enmeshed with concerns about the power, profit and motivations of 'big tech'. If the Government want to expand the adoption of AI into public services, they need to reassure the public that decisions about design, adoption, and data use are in the public interest. That requires shining a light on where and how the private sector are engaged in shaping public sector decisions. In our new report Licence to Build, we recommend the Science, Innovation and Technology initiative an inquiry into private technology companies' involvement in shaping policy and procurement decisions in public sector AI. Clarity and scrutiny will be essential to build public trust, and greater transparency will be a key first step.

Taxes or fees imposed on AI developers or deployers could shape market behaviour and improve the resourcing of existing regulatory systems. Cristina Muresan's new post on the Ada blog explores AI levies, drawing on the ‘polluter pays’ principle of environmental policy. The blog post looks at four historical and contemporary case studies from other industries to discuss key lessons for AI levies and raise questions for future research. https://lnkd.in/efn-g84c