Friedrich Naumann Foundation

Getting all the stakeholders together and setting out a clear mandate a manage expectations is an important first step for the first session in Geneva. Ultimately, success will be defined by finding consensus on governance and policy issues. It will be difficult, if not impossible to find consensus, given the geopolitical value that AI technology holds. Capacity building initiatives on the one side and regulation on the other side should be considered successful outcomes.

• Protection and promotion of human rights
• Transparency, accountability, and human oversight
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Open-source software, open data and open AI models

The promotion and protection of human rights should be inherent to the lifecycle of every technology, including AI. Transparency, accountability and human oversight are therefore a logical safeguard within the lifecycle of the technology. At no point, and especially not in these very early and vulnerable systems, should there be a situation where AI is deployed without human oversight. Social, economic, ethical, cultural, linguistic and technical implication of AI should also be a priority. As we have seen with the recent deterioration of the social media environment - which raises health concerns, including mental health, social isolation and addiction, these technologies pose a potential threat to society at large. In a capacity building perspective, open source software, open data and open AI models are a must to ensure that states without the capabilities to lead the development of these technologies can catch up, understand and learn the technology while it is being deployed - truly without being left behind. Additionally it enables individuals to benefit from technological advancements which in turn will foster innovation globally.

While implicated in 'Safe, secure and trustworthy AI', the cybersecurity aspect is a bit too absent in my opinion. AI generates huge amount of data and data flows. This exposes more data than ever before to the risk of being surveilled, captured and monetized. The use of AI in cybersecurity must be addressed imminently - as highlighted by Anthropics Mythos being withheld from the public due to its cyber capabilities. These kinds of security issues must be discussed with all stakeholders. While Anthropic seemingly 'did the right thing' in this case, these decisions should not only be left to the developers since it can have dire implications for the global internet infrastructure.

Data is leaking everywhere. The big AI companies such as Google and OpenAI that offer their models mainly through the cloud gather an immense breadth of data about the users personal lives, interests, situations, hardships, mental health state etc. On the other hand, they also gather lots of confidential data from businesses. Either due to lack of internal regulation and training or to be more efficient, many employees use AI in the workplace and feed it sensible information. There shall be guidelines and safeguards to prevent companies like Alphabet to use this data for more targeted ads or to let a state access it for 'national security' reasons.

The Dialogue can and shall function as the platform where governance ideas, challenges, differences and opportunities as perceived by all the stakeholders shall be brought up and discussed. In the best case, these discussions are as free as possible from politicization as possible in order to remain effective. The AI dialogue shall figure out the direction where AI Governance shall be going globally and serve as a forum for discussion while remaining impartial and based on scientific facts.

The AI Dialogue should coordinate with the Internet Governance Forum as it has a similar institutional role and has been around for quite a while now. Also, since the Dialogue emphasizes multistakeholderism, it should coordinate with ICANN. Thematically they are not the closely related, at least not as close as the Dialogue is to the IGF, but since ICANN is an important non governmental organization, the Dialogue should consult with them. In a perfect world the AI Dialogue partners up with the most important AI private sector companies such as OpenAI and Google in order to involve them heavily in the discussion from the beginning. However, a balance must be struck between not transforming the Dialogue into a lobby event for big tech and keeping them interested and involved in the discussion.

Thematic breakout sessions in smaller groups. Q&A with experts taking questions from any stakeholder.

Women in general and the global south as a whole are underrepresented in global discussion on AI governance. The reasons for their exclusion may vary but getting them to engage and be included in these discussion, even if they are not as imminently affected by the outcomes of the discussions, is of utmost importance for a balanced discussion within the AI Dialogue. Women are generally most affected by the misuse of AI on an individual level and therefore need to be more represented in Governance discussions to help discuss and set the guidelines and framework for policy related to the use of AI.

Thematic breakout sessions in smaller groups. Q&A with experts taking questions from any stakeholder.

Open source models and research promotes effective AI governance and supports individual freedoms such as the right to privacy by allowing the models to be run locally. Not releasing dangerous models such as the Mythos from Anthropic is generally a good practice but this should be more strictly regulated and there should be oversight over the public release of new models.