Articles tagged with #prompt-injection
An AI agent echoed a secret key, token, or connection string in its response or a tool call. Rotate it in minutes, audit for misuse, and block the model from ever seeing raw secrets again.
A user tells your AI to play a character with 'no restrictions' and it produces policy-violating output. Detect roleplay filter bypass in logs and add output-side guardrails that actually hold.
An uploaded file carries hidden instructions that hijack the AI mid-task. Detect white-text, Unicode-smuggled, and metadata payloads, sanitize uploads, and block file-triggered tool calls.
Your AI content tool produced a convincing phishing email or credential-harvesting page because the request was reframed as marketing or training. How to detect the three-signal pattern and add output-side intent gates.
An AI agent encodes sensitive data into a Markdown image URL and the chat UI auto-loads it, leaking data to an attacker's server. How to detect the gadget and break the chain.
A PDF carries invisible white-on-white, tiny-font, or metadata text that overrides your AI pipeline. Detect, strip, and harden against PDF-borne indirect prompt injection.
An AI agent fetches a URL and the page's hidden text hijacks its next action. Detect and block indirect prompt injection from web content.
An AI agent calls send_email or a webhook right after a web search. The fix: scan and delimit every snippet as untrusted data, and gate side-effecting tools behind a search boundary.
Hidden instructions ride into your AI pipeline through a translation step — either the LLM translator executes them, or a translation API's output re-enters unscanned. Detection, scanning, and isolation fixes.
A trusted MCP package ships a malicious update that steals keys or makes unexpected tool calls. Detect the compromise, contain it, rotate secrets, and harden the install with version pinning and a release-age cooldown.
An attacker shifts model behavior one message at a time until restrictions fall. Detect the escalation pattern across the conversation, reset context, and add session-level monitoring that fires before the violation turn.
A crafted user message overrides the system-prompt policy and the model ignores its guardrails. Diagnose which bypass bucket you are in, then build a layered defense that holds.
A code comment told your AI reviewer to approve a PR or skip a section, and it obeyed. How to detect comment injection, label code as data, and shrink the blast radius.
An uploaded file's name carries AI override instructions that fire when the agent reads the filename. The fix: reference files by UUID in prompts, never the raw name. Detect, sanitize, and block filename-borne injection.
User-pasted text smuggles override instructions that hijack your AI assistant. Detect the trust-boundary break and harden your app so pasted content can't act as instruction.
An API key, password, or token ended up inside a prompt. How to rotate it fast, trace where it traveled, and stop secrets from entering the model context again.
A user convinces your AI it now has system-level authority. Spot role confusion in logs and stop it with structural trust-tier enforcement, not better wording.
An agent runs a tool, the result contains hidden instructions, and the model obeys them. Why tool output gets user-level trust, and the role, labeling, and capability-gate fixes that stop it.
A rogue MCP server hides instructions in a tool's description so the model exfiltrates data while running a normal tool. Detect, audit, and harden against tool poisoning.
User-typed text lands in the system/developer role or operator-trust position, so the model obeys it like a developer directive. Root causes and the architectural fix, verified June 2026.