Lawyers
Cybersecurity & Data Protection Practice Group Chair John Carlin spoke with Lexology about the need for companies to update artificial intelligence governance and policies in response to the increased use of AI agents. In “The rise of autonomous agents: why your AI policies are no longer fit for purpose,” published on March 19, John says that businesses must anticipate and test the behavior of agents. “Sandboxing is a great governance and risk management measure, allowing a business to identify potential adverse impacts of an agentic AI system before it is fully deployed,” John says.
He adds that policy is just one part of mitigating AI risks. Employees must also be trained to recognize risks such as “prompt injections: which are malicious hidden instructions that hijack agent behaviour,” John notes.
» read the article
Related Insights
