AI Agents Lack Established Principles, Panelists Say
At the Youth Internet Governance Forum India, experts highlighted the absence of specific principles and clear liability frameworks for artificial intelligence agents. The discussion focused on cybersecurity challenges in the AI era.
There are no established principles specifically for AI agents, according to Gowree Gokhale, an independent lawyer, speaking after a panel on 'Cybersecurity in the Age of AI' at the Youth Internet Governance Forum India (IGF).
The event, held at IIT Bombay on September 4, 2026, featured a hypothetical scenario where autonomous AI agents could breach a small nation, prompting discussions on accountability. This scenario was highlighted by recent incidents, such as the OpenAI-Hugging Face event, demonstrating the unpredictable nature of AI agents.
Gokhale pointed out that existing intermediary liability frameworks, like those under India's IT Act, may not be sufficient for autonomous AI agents. These agents can initiate transmissions, select recipients, and modify information by design, diverging from the conditions for safe harbor protections.
Participants, divided into government, private sector, civil society, and technical community groups, explored potential solutions. The need for a due-diligence regime and clear lines of responsibility was emphasized, particularly in light of AI agents' potentially unpredictable actions.
Experts suggested that managing AI agents might require strategies similar to those used for child development, focusing on guidance and understanding their inherent tendencies rather than solely relying on prohibitions.