Yvette Khozam, ePMA lead pharmacist at West London NHS Trust, highlights critical gaps in safety standards for AI in healthcare. Despite NHS England’s aim to establish the NHS as the leading AI-enabled care system, current frameworks were last updated in 2018, failing to address the complexities of modern AI technology. Existing clinical risk standards, designed for predictable software, overlook AI’s adaptive capabilities, leading to potential risks like “in-context deception” and steganography. Khozam stresses the urgency for three immediate actions: ensuring transparency in AI systems, conducting thorough tests for hidden behaviors, and monitoring for semantic drift. As adaptive AI is already integrated into clinical settings, she emphasizes the importance of aligning safety standards with these technologies. With initiatives underway, such as updated regulations and AI safety systems, Khozam advocates for proactive governance to mitigate risks and harness AI’s potential effectively in healthcare.
Source link
Share
Read more