Prompt injection
23 moments from 20 videos · 1:12:27 total
Learn how to mitigate security risks in language models. These conference talk segments explain how prompt injection attacks occur and show how to build resilient defensive layers.
Security risks involving prompt injection attacks
How prompt injections allow users to manipulate model outputs similar to SQL injection attacks.
Exploiting systems through audio manipulation and prompt injections
Unconventional security vulnerabilities like hidden text in applications and audio-induced system crashes showcase the unintended consequences of digital processing.
Prompt injections bypassing language model service guardrails
Indirect prompt injections can turn well-meaning language models into accomplices that generate executable malicious scripts.
Testing edge cases and mitigating prompt injection attacks
Designing custom tests for bad actors prevents prompt injection by routing malicious manipulations to a secure human fallback.
Live demo: Stopping prompt injections dynamically
Utilizing deterministic task binding and policy to prevent unauthorized actions triggered by malicious inputs.
Evaluating prompt injection risks in compliant enterprise environments
Deploying unmonitored agents poses severe infrastructural threats if prompt injections alter critical system configurations.
Identifying and defining prompt injection execution vulnerabilities
Malicious language input bypasses instructions by treating natural language text fields as hidden executable commands.
Executing basic prompt injection attacks through translation tasks
Conflicting instructions in user input can successfully override a base language model's original operational directive.
Comparing prompt injection to traditional SQL injection attacks
The fundamental lack of separation between developer code and user input in language models mirrors the mechanics of SQL injection.
Mitigating prompt injection vulnerabilities in external AI tool integrations
Restricting external data permissions neutralizes extreme security vulnerabilities like automated prompt injections found in active agent protocols.
Prompt injection vulnerabilities and contextual mitigation testing challenges
Why complex contextual prompts easily circumvent manual ethical constraints and scale poorly for foundational models.
Prompt injection and data exfiltration inside IDEs
How malicious repository comments manipulate coding assistants to leak sensitive developer data.
Protecting systems from prompt injection and sentiment attacks
Guarding against prompt manipulation and negative sentiment inputs prevents unintended model overrides and resource starvation.
Differences between direct and indirect prompt injections
How attackers manipulate model outputs using explicit text inputs or hidden instructions within multimedia formats.
Mitigating prompt injections using guardrails and filters
Applying human oversight, capability restrictions, and output filtering mechanisms to control malicious prompt instructions.
Drone prompt injection vulnerabilities and rapid 3D city mapping
Physical road signs expose autonomous drones to prompt injection vulnerabilities alongside theoretical capabilities for rapid urban mapping.
Securing environments against prompt injection and key exposure
Implementing strict safeguards for environment variables prevents accidental execution costs and database exposure via language models.
Combating prompt injection in automated applicant tracking systems
The massive influx of generated applications requires recruiters to identify hidden instructions and consistently verify candidate authenticity.
Identifying common and emerging application injection attack vectors
Understanding why vulnerabilities ranging from legacy database queries to modern text prompt manipulations persistently occur.
Prompt injections and extracting locked API keys from LLMs
Bad actors exploit indexed linguistic models using jailbreaks and policy puppetry to harvest sensitive environment variables.
Deploying AI firewalls to prevent prompt injections by non-technical staff
Utilizing advanced AI firewalls and dynamic execution blocking intercepts data leaks prompted by unaware business users.
Preventing document pipeline exploitation and data poisoning
Scrubbing internal deposits against potential prompt injection payloads halts unchecked malicious execution capabilities.
Q&A on platform discovery, security, and prompt injection
Resolving challenges related to AI engine optimization, protecting endpoints from malicious payloads, and tracking framework updates.
Your mix. Instantly.
More mixes
AI & Machine Learning
AI coding agents
AI & Machine Learning
MCP (Model Context Protocol)
AI & Machine Learning
RAG
Languages & Frameworks
TypeScript
Languages & Frameworks
React
Languages & Frameworks
Java
Cloud, Platform & Data
Platform engineering
Cloud, Platform & Data
Observability
Cloud, Platform & Data
Infrastructure as code
Architecture, Quality & Security
System architecture
Architecture, Quality & Security
Code review
Architecture, Quality & Security
Software supply chain security
Leadership, Career & Community
Engineering strategy
Leadership, Career & Community
Developer experience
Leadership, Career & Community