F5 AI Guardrails deploys as a proxy between customers and AI fashions. Wormke describes it as being inserted as a proxy layer on the “entrance door” of AI interplay, between AI purposes, customers and brokers. It intercepts prompts earlier than they attain the mannequin and analyzes outputs earlier than they return to customers. The system applies coverage guidelines to precise content material relatively than transport-layer traits.
Coverage enforcement covers a number of classes. Guardrails blocks prompts that try jailbreaks or injection assaults, scans outputs for delicate knowledge patterns, and enforces compliance necessities together with GDPR and the EU AI Act.
Crimson Workforce automation creates steady testing
AI Crimson Workforce automates adversarial testing towards AI methods. It maintains a database of assault methods that grows by 10,000 entries month-to-month as researchers uncover new vulnerabilities.
Outcomes from Crimson Workforce testing feed instantly into Guardrails insurance policies. When Crimson Workforce discovers a vulnerability sample, safety groups create corresponding guardrails to dam comparable assaults in manufacturing.
“It’s a really synergistic pairing whereby with AI Crimson Workforce you possibly can ship a crew of brokers to find vulnerabilities in AI methods, and with AI Guardrails you possibly can rework these insights into threat-informed defenses,” Wormke stated.
NGINXaaS delivers enterprise load balancing as a managed service
F5 acquired NGINX again in 2019 and has been increasing the capabilities of the online server platform ever since.
