#prompt injection
87 stories taggedprompt injection · page 5 of 6.
CrowdStrike Flags Five New Ways Criminals Can Trick AI Into Obeying Them
Researchers have mapped out a fresh set of prompt injection attacks, where criminals feed deceptive instructions to AI systems to make them behave in harmful ways. Here is what each one does and why ordinary employees are part of the risk picture.
AI Coding Assistants Can Be Tricked Into Running the Very Malware They Were Asked to Find
A proof-of-concept from the AI Now Institute shows Claude Code and OpenAI's Codex executing attacker-supplied code when asked to review it in autonomous mode.
A Hidden Note in a Bug Report Tricked GitHub's AI Into Leaking Company Secrets
Researchers showed how a single crafted message in a public GitHub issue could fool an AI assistant into reading private code and posting it online for anyone to see.
Researchers Show How a Fake GitHub Comment Can Trick AI Tools Into Leaking Secret Code
A crafted public comment on GitHub can manipulate AI-powered automation into handing over data from private repositories, no password required.
A Hidden Command in a GitHub Issue Can Silently Steal a Company's Private Code
Researchers found a flaw in GitHub's AI automation tool that lets an outsider read an organisation's private repositories by hiding plain-English instructions inside a public bug report.
AI Agents Can Be Tricked Into Sending Money. Zscaler Has the Data.
A new study shows that some expensive, enterprise-grade AI assistants fall for hidden instructions that most humans would ignore, and the real danger is far bigger than a fake three-dollar fee.
The Weak Link This Week Wasn't Code. It Was Trust.
From home streaming boxes turned into criminal relays to AI assistants tricked by hidden instructions, this week's incidents share one root cause: systems trusting the wrong thing.
Hackers Are Hiding Instructions Inside Websites to Make AI Assistants Send Crypto Payments
Two newly discovered attack campaigns show how criminals can secretly hijack AI browsing agents by planting hidden commands in ordinary-looking web pages.
Popular AI Coding Tool Cursor Has Flaws That Could Let Attackers Run Code on Your Computer
Security researchers found two vulnerabilities in the Cursor AI code editor that could allow an attacker to silently take control of a developer's machine, no click required.
A Fake Error Message Hijacked AI Coding Assistants — and Security Tools Saw Nothing
Researchers planted a single bogus bug report in a popular developer service and watched AI coding agents obediently run the attackers' code. No password stolen. No alarm raised.
Context Manipulation Attack 'BioShocking' Turns Agentic Browsers Into Credential Thieves
Researchers show how poisoned context fed to AI-driven browser agents causes them to drop safety guardrails and quietly exfiltrate stored credentials.
Cursor IDE's Sandbox Cracked by Prompt Injection — No User Interaction Required
Two logic flaws in Cursor's command execution sandbox let attackers escape the isolation layer and run code on the underlying OS. Patches landed in April. The researchers say Cursor isn't alone.
DuneSlide: Two Cursor Bugs Turn a Prompt Into a Shell
A pair of 9.8-rated flaws in the AI code editor let a single crafted prompt escape the sandbox and execute arbitrary commands, no user approval required.
Poisoned Tool Descriptions Turn Helpful AI Agents Into Quiet Exfiltration Channels
Microsoft Incident Response shows how a single malicious MCP tool description can coax an agent into leaking corporate data without tripping a single policy check.
BioShocking: Prompt-Game Trick Pries Credentials From AI Browsers
Researchers at LayerX got six AI browsers and assistants, including ChatGPT Atlas, Perplexity's Comet and Anthropic's Claude extension, to exfiltrate user logins by framing the attack as a game.