Amre Metwally

This presentation from Anthropic, will delve into the forefront of agentic AI safety. Exploring the critical work that goes into robust policy frameworks and reliable mechanisms to manage autonomous systems, discussing the complexities of new model deployment and strategies to counter misuse. Shaping the next wave of safe agentic AI through both technical safeguards and proactive engagement with public policy and regulators.

• Building consent frameworks that empower users and ensure transparent, manageable interactions with increasingly autonomous AI.
• Policy development and governance structures to ensure the safe and responsible creation and deployment of AI models.
• Strategies for safeguarding against "jailbreaks" and maintaining robust control over product behaviour in our newest AI models.
• Collaborating with policymakers to foster a responsible and safe environment for advanced AI development and deployment.
• Exploring how advanced AI, including agentic capabilities, can fundamentally change T&S dynamics, leading to more sophisticated automation and threat detection.