Securing AI models and using AI to maintain compliance
summary
IBM demonstrates an alpha service called Autonomous Security and Compliance (ASC) that runs on AWS Bedrock to ingest customer policies via RAG, generate and deploy CloudFormation templates, monitor AWS Config/Security Hub/GuardDuty for drift, and take staged remedial actions from monitor-only to active remediation. The stack includes AI firewalls, model scanning, runtime monitoring and adversarial testing to catch prompt injection, poisoning or leakage, and a service loop for fine‑tuning models as incidents and compliance gaps are observed.
AI-written summary · about
chapters (7)
- 0:00Why securing AI matters
- 4:41AI security framework overview
- 6:49Securing AI models and data
- 9:20AI security tools and threats
- 11:32Using AI for cybersecurity
- 13:21Autonomous Security and Compliance
- 16:56IBM services and Q&A
related talks
28mSecuring AI by design
40mEnsuring Data Security in the AI Revolution
12mBuilding Resilient AI: A Framework for Enterprise Security
22mHow IBM enhances security on AWS and your hybrid cloud
29mArchitect AI success in a regulated world
12mUsing AI/ML to scale governance, risk management, and audits
15mAI Guardrails for Enterprise Deployments
08mshares insights on AI and regulatory compliance