> Markdown version of [/playlists/prompt-injection](https://www.wearedevelopers.com/playlists/prompt-injection). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md). --- # Playlist: Prompt injection 20 videos · 23 moments · 72.5 minutes ## The shadows that follow the AI generative models - **Security risks involving prompt injection attacks** (14:26, 1min) — How prompt injections allow users to manipulate model outputs similar to SQL injection attacks. [Learn more](https://www.wearedevelopers.com/videos/624-the-shadows-that-follow-the-ai-generative-models) ## WeAreDevelopers LIVE - Is Software Ever Truly Accessible? - **Exploiting systems through audio manipulation and prompt injections** (33:46, 5min) — Unconventional security vulnerabilities like hidden text in applications and audio-induced system crashes showcase the unintended consequences of d... [Learn more](https://www.wearedevelopers.com/videos/1361-wearedevelopers-live-is-software-ever-truly-accessible) ## Dangerous Reactivity: Why AI Output Is the New XSS - **Prompt injections bypassing language model service guardrails** (06:02, 3min) — Indirect prompt injections can turn well-meaning language models into accomplices that generate executable malicious scripts. [Learn more](https://www.wearedevelopers.com/videos/100115-dangerous-reactivity-why-ai-output-is-the-new-xss) ## The AI Agent Path to Prod: Building for Reliability - **Testing edge cases and mitigating prompt injection attacks** (19:26, 4min) — Designing custom tests for bad actors prevents prompt injection by routing malicious manipulations to a secure human fallback. [Learn more](https://www.wearedevelopers.com/videos/1523-the-ai-agent-path-to-prod-building-for-reliability) ## The day the chatbot asked for sudo - **Live demo: Stopping prompt injections dynamically** (20:05, 2min) — Utilizing deterministic task binding and policy to prevent unauthorized actions triggered by malicious inputs. [Learn more](https://www.wearedevelopers.com/videos/100038-the-day-the-chatbot-asked-for-sudo) ## Unlocking Value from Data: The Key to Smarter Business Decisions- - **Evaluating prompt injection risks in compliant enterprise environments** (12:30, 1min) — Deploying unmonitored agents poses severe infrastructural threats if prompt injections alter critical system configurations. [Learn more](https://www.wearedevelopers.com/videos/1367-unlocking-value-from-data-the-key-to-smarter-business-decisions) ## The AI Security Survival Guide: Practical Advice for Stressed-Out Developers - **Identifying and defining prompt injection execution vulnerabilities** (06:05, 3min) — Malicious language input bypasses instructions by treating natural language text fields as hidden executable commands. [Learn more](https://www.wearedevelopers.com/videos/1015-the-ai-security-survival-guide-practical-advice-for-stressed-out-developers) ## ChatGPT, ignore the above instructions! Prompt injection attacks and how to avoid them. - **Executing basic prompt injection attacks through translation tasks** (02:05, 1min) — Conflicting instructions in user input can successfully override a base language model's original operational directive. - **Comparing prompt injection to traditional SQL injection attacks** (13:04, 2min) — The fundamental lack of separation between developer code and user input in language models mirrors the mechanics of SQL injection. [Learn more](https://www.wearedevelopers.com/videos/723-chatgpt-ignore-the-above-instructions-prompt-injection-attacks-and-how-to-avoid-them) ## Three years of putting LLMs into Software - Lessons learned - **Mitigating prompt injection vulnerabilities in external AI tool integrations** (24:53, 1min) — Restricting external data permissions neutralizes extreme security vulnerabilities like automated prompt injections found in active agent protocols. [Learn more](https://www.wearedevelopers.com/videos/1508-three-years-of-putting-llms-into-software-lessons-learned) ## Hacking AI - how attackers impose their will on AI - **Prompt injection vulnerabilities and contextual mitigation testing challenges** (19:14, 5min) — Why complex contextual prompts easily circumvent manual ethical constraints and scale poorly for foundational models. [Learn more](https://www.wearedevelopers.com/videos/824-hacking-ai-how-attackers-impose-their-will-on-ai) ## Can Machines Dream of Secure Code? Emerging AI Security Risks in LLM-driven Developer Tools - **Prompt injection and data exfiltration inside IDEs** (31:25, 1min) — How malicious repository comments manipulate coding assistants to leak sensitive developer data. [Learn more](https://www.wearedevelopers.com/videos/1217-can-machines-dream-of-secure-code-emerging-ai-security-risks-in-llm-driven-developer-tools) ## Beyond the Hype: Building Trustworthy and Reliable LLM Applications with Guardrails - **Protecting systems from prompt injection and sentiment attacks** (12:11, 2min) — Guarding against prompt manipulation and negative sentiment inputs prevents unintended model overrides and resource starvation. [Learn more](https://www.wearedevelopers.com/videos/1594-beyond-the-hype-building-trustworthy-and-reliable-llm-applications-with-guardrails) ## Prompt Injection, Poisoning & More: The Dark Side of LLMs - **Differences between direct and indirect prompt injections** (02:27, 2min) — How attackers manipulate model outputs using explicit text inputs or hidden instructions within multimedia formats. - **Mitigating prompt injections using guardrails and filters** (04:46, 2min) — Applying human oversight, capability restrictions, and output filtering mechanisms to control malicious prompt instructions. [Learn more](https://www.wearedevelopers.com/videos/1563-prompt-injection-poisoning-more-the-dark-side-of-llms) ## AI Space Factories, Hacking Self-Driving Cars & Detecting Deepfakes - **Drone prompt injection vulnerabilities and rapid 3D city mapping** (02:55, 0min) — Physical road signs expose autonomous drones to prompt injection vulnerabilities alongside theoretical capabilities for rapid urban mapping. [Learn more](https://www.wearedevelopers.com/videos/1812-ai-space-factories-hacking-self-driving-cars-detecting-deepfakes) ## Slopquatting, API Keys, Fun with Fonts, Recruiters vs AI and more - The Best of LIVE 2025 - Part 2 - **Securing environments against prompt injection and key exposure** (06:14, 3min) — Implementing strict safeguards for environment variables prevents accidental execution costs and database exposure via language models. - **Combating prompt injection in automated applicant tracking systems** (43:46, 8min) — The massive influx of generated applications requires recruiters to identify hidden instructions and consistently verify candidate authenticity. [Learn more](https://www.wearedevelopers.com/videos/1765-slopquatting-api-keys-fun-with-fonts-recruiters-vs-ai-and-more-the-best-of-live-2025-part-2) ## How to Defend Against Data Manipulation Attacks - Bozidar Spirovski & Wekoslav Stefanovski - **Identifying common and emerging application injection attack vectors** (02:31, 2min) — Understanding why vulnerabilities ranging from legacy database queries to modern text prompt manipulations persistently occur. [Learn more](https://www.wearedevelopers.com/videos/1829-how-to-defend-against-data-manipulation-attacks-bozidar-spirovski-wekoslav-stefanovski) ## WeAreDevelopers LIVE - Chrome for Sale? Comet - the upcoming perplexity browser Stealing and leaking - **Prompt injections and extracting locked API keys from LLMs** (15:09, 4min) — Bad actors exploit indexed linguistic models using jailbreaks and policy puppetry to harvest sensitive environment variables. [Learn more](https://www.wearedevelopers.com/videos/1331-wearedevelopers-live-chrome-for-sale-comet-the-upcoming-perplexity-browser-stealing-and-leaking) ## The New AI Security Stack: Observe, Detect, Protect - **Deploying AI firewalls to prevent prompt injections by non-technical staff** (21:52, 2min) — Utilizing advanced AI firewalls and dynamic execution blocking intercepts data leaks prompted by unaware business users. [Learn more](https://www.wearedevelopers.com/videos/100302-the-new-ai-security-stack-observe-detect-protect) ## RAG like a hero with Docling - **Preventing document pipeline exploitation and data poisoning** (17:19, 2min) — Scrubbing internal deposits against potential prompt injection payloads halts unchecked malicious execution capabilities. [Learn more](https://www.wearedevelopers.com/videos/1602-rag-like-a-hero-with-docling) ## API, MCP or MCP App? Choosing the right surface for AI agents - **Q&A on platform discovery, security, and prompt injection** (25:02, 4min) — Resolving challenges related to AI engine optimization, protecting endpoints from malicious payloads, and tracking framework updates. [Learn more](https://www.wearedevelopers.com/videos/100305-api-mcp-or-mcp-app-choosing-the-right-surface-for-ai-agents)