The widespread deployment of artificial intelligence (AI) has seen a renewed imperative for industry and its leaders, or “technocrats”, to focus on enhancing trust in varied applications of AI. The rapid prominence of AI has rejuvenated a reliance on experts to navigate AI’s complexities and harness its potential for social good. It is increasingly crucial that developers, users, and public institutions—both civil society and the government—join forces to ensure the trustworthy design, development, and deployment of AI applications. Only through this inclusive approach can we build AI that truly benefits everyone.

Currently, AI research remains predominantly driven by industry, which plays a leading role in shaping the technology’s trajectory. Industry pioneers enjoy essential inputs such as access to vast datasets, top-tier talent, large-scale models, and computing power. Industry accounts for 96% of the largest AI models  in any given year, the size of models measured in terms of parameters for supervised, unsupervised, and reinforcement learning models. In a telling comparison, the industry has produced 51 notable machine learning models, while academia has produced only 15. The count includes models selected by the research institute Epoch AI based on their historical significance, number of citations, and state-of-the-art advancement. Given this reality, industry stewardship is indispensable for building a trustworthy AI ecosystem in the near and medium term.

Policymakers and regulators are increasingly collaborating with industry to understand and steer AI design-development-deployment cycles to ensure safety and trustworthiness. The initiatives span across the multi-lateral, national, and sector-specific level, embodying a collective desire to establish guardrails.  

In terms of multi-lateral efforts and co-ordination, this dynamic was evident at the inaugural AI Safety Summit in Bletchley and its successor in Seoul. The conscientious collaboration between government and industry has led to the interim AI spearheaded by Turing Award winner Prof. Yoshua Bengio. The report underscores the necessity for international collaboration to enhance understanding and implement safety measures for general-purpose AI, stating that “amid rapid advancements, research on general-purpose AI is currently in a time of scientific discovery and is not yet settled science.” Additionally, the UN released its first report titled Governing AI for Humanity to meet the need for global collaboration, transparency, and ethical guidelines to ensure globally shared benefits, mitigate risks, and promote inclusive development. Over 60 countries are utilizing UNESCOs Readiness Assessment Methodology to help states identify their level of AI technological readiness and implement policies accordingly. Even the OECD has updated its principles on AI to better safeguard information integrity against the potential adverse effects of Generative AI (Read more on our blog). Attempting to implement the OECD principles, the Global Partnership on Artificial Intelligence (GPAI) has established four working groups on data governance, future of work, innovation and commercialization, and responsible AI. It is extremely encouraging to witness efforts towards proactive collaboration for defining scope and modalities in AI governance at the multilateral level. This is essential for harmonizing regulatory frameworks and addressing the complex ethical and societal implications of AI technologies on a global scale.

On the national level of AI governance, the OECD AI Policy Observatory documents 69 countries with AI laws, representing a significant increase over the past few years due to a strategic evolution in AI governance according to research conducted by Deloitte. AI regulation has expanded in three phases: understanding AI, fostering its growth, and actively shaping its development. Nation-States are establishing national AI Safety   institutes as seen in UK, the US and Singapore. In the US, the Senate has hosted several AI “listening sessions” with representatives from BigTech, underscoring the reemergent central role of technocrats in fostering critical discussions. In the EU, partnerships with AI companies are focusing on projects like the AI4Cities initiative, which aims to use AI to reduce carbon emissions in urban areas. In Japan, collaborations with AI firms are addressing the aging population crisis through developments such as AI-powered robots to assist with elder care and smart healthcare solutions to improve medical services. National-level AI collaborations are pivotal in addressing country-specific challenges, fostering a conducive innovation ecosystem, and ensuring that the benefits of AI advancements are equitably distributed across various sectors.

Initiatives to ensure the safe and ethical advancement of AI also emerge from industry. A whistleblower letter garnered support across multiple companies, and an open letter documented industry-wide concerns about AI’s potential threats. At the same time, leading AI firms are publishing principles and best-practices to ensure AI safety. Microsoft has offered insight into their red teaming practices and Anthropic is publishing their policy vulnerability tests. Partnerships and coalitions have also sprung up to face the novel challenges of AI. Adobe established a coalition for content provenance and the Partnership on AI – a consortium of industry, government, and civil society organizations – seeks to ensure the authenticity and transparency of digital content, combat misinformation, and promote ethical AI practices through shared standards for content attribution and provenance. Meta also posted their results of a Community Forum survey in collaboration with Stanford, which revealed a strong public demand for increased transparency in AI development, ethical guidelines to prevent misuse, and greater accountability for AI systems. Tech companies are increasingly collaborating with each other and academia to diversify their engagement with civil groups, fostering inclusive dialogue and ensuring AI development aligns with societal needs and values.

Despite the growing collaborations across multiple levels, the risks and potential harms of AI continue to rise. The OECD’s AI Monitor tracks a growing number of incidents and harms over the past year. AI tools have been misused to spread misinformation, for example during the EU elections when an AI-generated video impersonating an official claimed to have rigged the elections. Other forms of AI misuse are also well-documented, including scams and fraud, harassment, and opinion manipulations. Additionally, lacking infrastructure and deep inequalities limit the potential of AI as a force for good. International Telecommunication Union estimates that 2.6 billion people are still offline, unable to reap the benefits from AI. The continuously evolving AI technologies and applications have thus far outpaced the efforts to harness their benefit safely, responsibly, and equitably.

The global commons have taken cognizance of dealing with the ascent of AI, both by mitigating its risks and distributing its benefits. However, the full comprehension of this aspiration relies on a trustworthy ecosystem that promotes ethics, diversity of perspective, sustainability, equity, accountability, and transparency. Herein, industry plays a critical role, but civil society, governments, and the publics should be elevated to equal statues of importance. Together, they must become stakeholders in setting a trustworthy foundation for a technological disruption with the potential to improve or exacerbate prevalent suffering.