01
Risks:
Focuses on Foundation Models, App Intents.
AI & Machine Learning / 25:12
Secure your app: mitigate risks to agentic features
AI & Machine LearningPrivacy & Securityiosipadosmacostvosvisionoswatchos
Key Points
- Explains how to evaluate threats from indirect prompt injection, such as data exfiltration and unintended actions.
- Covers security practices for App Intents and Foundation Models, including user confirmations, secure prompt design, and mitigations for agentic workflows.
- The session moves through Risks, Threat modeling, Implementing mitigations, Foundation Models, App Intents.
- Key concepts include App Intents, Foundation Models, LLMs, Siri, PostAndFetchPublicFeedTool, LanguageModelSession, Apple Intelligence, Shortcuts.
- Platform coverage: ios, ipados, macos, tvos, visionos, watchos.
Condensed Flow
02
Threat modeling:
Focuses on PostAndFetchPublicFeedTool.
03
Foundation Models:
Focuses on LanguageModelSession.
04
App Intents:
Focuses on Apple Intelligence, Siri, Shortcuts.
More Detail
Additional details
- Detailed flow: Risks -> Threat modeling -> Implementing mitigations -> Foundation Models -> App Intents.
- APIs and concepts to recognize: App Intents, Foundation Models, LLMs, Siri, PostAndFetchPublicFeedTool, LanguageModelSession, Apple Intelligence, Shortcuts, OrderTeaTool, DeletePhotoIntent, AppIntent, Siri AI.
- Version and support notes focus on Siri.
- Implementation focus: Foundation Models, App Intents, OrderTeaTool, DeletePhotoIntent, LLMs, LanguageModelSession, Siri, AppIntent, PostAndFetchPublicFeedTool, Siri AI.