Singapore, Sept. 27, 2024 – At the AI Apex Asia Capital Connect Forum, a panel of leading experts in AI governance, security and alignment, intellectual property, and regulatory compliance convened to discuss the complex landscape facing AI companies preparing for initial public offerings (IPOs). The roundtable, moderated by James Liu, Director at Alibaba Cloud International, provided crucial insights into the evolving regulatory environment and strategies for success.
Key findings from the discussion include:
1. AI companies face unprecedented challenges in data privacy, intellectual property protection, and regulatory compliance when preparing for IPOs.
2. The rapid pace of AI innovation often outstrips regulatory frameworks, requiring companies to be proactive in addressing potential risks and communicating with stakeholders effectively.
3. Collaboration between AI companies and regulators is crucial to develop frameworks that balance innovation with public interest protection.
4. The EU AI Act represents a significant shift in AI regulation, with global implications for companies operating in or selling to European markets.
5. Deepfake technology poses significant risks to copyright and identity protection, necessitating both technological and regulatory solutions.
6. Purpose-built AI agents offer a promising approach to augment Large Language Models (LLMs) for enhanced security and safety.
Prof. Liu Yang, Co-founder of AgentLayer, highlighted the evolving nature of AI security risks and proposed an innovative solution: “When you are using any AI solutions, you need to understand potential new attacks. For example, we’ve seen recent breakthroughs in AI model jailbreaking, which can bypass defensive mechanisms with 100% success for current models. This brings a very big challenge for whatever AI solutions we’re developing.”
Prof. Liu further elaborated on a novel approach to address these challenges: “Augmenting LLMs with agent-based models as verifier and governor recognizes the complexity of an enterprise while focusing on creating an automated LLM verifier and governor. Instead of focusing solely on making LLMs safe for an enterprise, purpose-built AI agents can be used to verify and govern all LLM interactions. Even if the underlying LLMs are not inherently safe, the use of these specialized agents can help an enterprise ensure its interactions are safe as well as outcome-oriented.”
Hsu Li-Chuan, Partner at Dentons Rodyk, emphasized the importance of clear communication with regulators and investors: “The difficulty with AI is it covers so many different areas of regulation, unlike e-commerce or even blockchain technology. AI probably encompasses a much wider area of regulation, compliance, and ethics. We need to keep this in mind when doing any fundraising on public markets.”
Yang Jingwei, Director of Ant Digital Technologies, stressed the need for robust technical solutions: “For data security, I have three suggestions: build very strong data governance with strict policies, be very careful with cross-border data transfers, and be transparent about data usage. For IP rights, consult with IP experts, seek patent protections, and have clear contracts for IP ownership.”
Yang Jingwei then mentioned the risk of Deepfakes highlighted in the EU AI Act and how to prevent it: “Deepfake attacks pose significant risks, but content labeling could help, though it requires time and investment. In the interim, companies are encouraged to partner with reliable vendors like ZOLOZ, and users are advised to seek secondary confirmation during high-risk activities like money transfers.”
James Liu, the moderator, concluded: “The key takeaway is that pre- and post-IPO, communication with investors, the general public, and regulators is crucial. Because technology often moves faster than regulation, self-regulation and frameworks are important. The innovative use of purpose-built AI agents to augment LLMs, as proposed by Prof. Liu, represents a promising direction for addressing current security and safety concerns in the AI industry.”
The panelists agreed that as AI continues to evolve rapidly, companies must stay ahead of governance issues to successfully navigate the IPO process and maintain public trust. Ongoing dialogue between industry, academia, and regulators will be crucial in shaping a balanced approach to AI governance, with emerging technologies like AI agents playing a key role in ensuring the safe and effective deployment of AI systems.
