Tag: prompt injection
58 articles
Malware Stunningly Evades AI in Critical npm Breach
Think your npm packages are safe? Researchers found a malicious npm package that talks to a remote AI-like controller, adapting at runtime to dodge scanners and quietly steal valuable data.
Prompt Injection Through Poetry: Exclusive Best Defenses
What if a poem could fool the guard? New research shows adversarial verse — and even $5 expired-domain hijacks — can cheaply and reliably bypass model guardrails, turning style and supply-chain trust into a dangerous new attack surface.
More Prompt||GTFO Exclusive Guide to Effortless Prompts
System prompts make AI assistants helpful — and can quietly turn them into persistent, data-harvesting agents. This guide explains how crafty instruction tweaks and PromptFix attacks corrupt the instruction stream and what to watch for to keep your assistant honest.
Prompt Injection: Exclusive Look at Dangerous AI Browsers
Think your AI assistant only reads whats on screen? Researchers warn that CometJacking — hidden prompts tucked into a URL — can trick “AI browsers” into handing over emails, calendar entries and cloud files without passwords or user prompts.
Microsoft Exclusive Warns of Dangerous Whisper Leak
Think encryption keeps your AI chats private? Microsoft warns that streaming language models can leak conversation topics through packet timing and size, letting a passive network observer turn traffic patterns into probabilistic guesses about what you said.
Multi-Turn Attacks Reveal Stunning Open-Weight LLM Flaws
What if the helpful chat that answers your questions could be slowly nudged into doing harm? Ciscos analysis shows multi-turn attacks can trick open-weight LLMs into unsafe or disallowed outputs—sometimes with success rates near 90%—putting search, support, education and other services at risk.
Gemini AI Exclusive: Dangerous Thinking Robot Malware
What if the AI meant to amplify our thinking could be turned into thinking robot malware that rewrites itself to hide from defenders? New research shows attackers chaining prompt- and log-injection tricks to weaponize Gemini into self-modifying, persistent surveillance agents that sidestep many standard safeguards.
Claude Desktop Extensions Exclusive: Critical Prompt Risk
Claude Desktop extensions make assistants truly useful — but when they execute local actions, attackers can turn innocent prompts into harmful commands. The recent command‑injection flaws in three extensions, now patched by Anthropic, are a reminder that convenience brings new security risks.
OpenAI Assistants API Exclusive: Critical SesameOp Backdoor
Imagine your helpful AI assistant secretly moonlighting as a command-and-control courier — researchers found the SesameOp backdoor using OpenAI’s Assistants API to stealthily ferry attacker commands and exfiltrated data. This clever pivot turns trusted productivity integrations into covert channels, forcing a rethink of how we govern and monitor AI tools.
Sneaky Mermaid attack: Exclusive Copilot data breach alert
A clever Sneaky Mermaid indirect prompt injection showed how hidden instructions buried in files could trick Microsoft 365 Copilot into leaking tenant data. Microsoft says it patched this specific flaw, but security teams warn the broader risk of stealthy, embedded prompt attacks is far from over.
Microsoft 365 Copilot Exclusive: Dangerous Mermaid Attack
The Mermaid attack revealed how a hidden prompt in an otherwise harmless file could trick Microsoft 365 Copilot into spilling emails and attachments. Microsoft patched the gap, but the episode is a clear reminder that giving AI broad access can turn convenience into a new, exploitable data risk.
Sneaky Mermaid attack: Exclusive critical Copilot leak
Researchers uncovered a Sneaky Mermaid trick that hid malicious instructions inside ordinary files to make Microsoft 365 Copilot leak tenant emails and attachments. Microsoft patched the specific vector, but the episode is a wake-up call about how AI assistants can be manipulated and why teams must shore up their digital defenses.
AI Vulnerability Reward Program: Exclusive $30K Best Win
Google’s new AI Vulnerability Reward Program offers up to $30,000 to researchers who responsibly report model flaws — a smart, practical move to incentivize fixes, curb abuse, and make AI safer for everyone.
CometJacking: Risky Attack Exposes Data — Must-See Fixes
One click can turn your helpful AI into a sneak thief — CometJacking hides malicious prompts in links that trick Perplexity’s Comet into leaking email, calendar and connected data. Stay safe by updating clients, reviewing agent permissions, and avoiding unfamiliar links while these agentic AIs get harder to fool.
log-to-prompt injection: Risky Gemini Flaw Exposed
Researchers uncovered three now-patched Gemini vulnerabilities that could let attackers use prompt- and log‑injection tricks to expose personal and corporate data — a stark reminder that AI conveniences like personalization and logging can become dangerous attack surfaces.
indirect prompt injection: Stunning Risk Exposed
A trio of vulnerabilities in Google’s Gemini shows how indirect prompt injection—hiding instructions in files, metadata or chained APIs—can trick AI into leaking data or taking unintended actions, proving that securing models means vetting every input source, not just user prompts.
prompt injection: Stunning $5 Domain Risk
Could a $5 expired domain let a stranger trick your AI into spilling customer data? Researchers proved it with Salesforce’s Agentforce, a wake-up call that mundane trust failures in AI pipelines can lead to serious leaks and that continuous domain monitoring and layered safeguards are essential.
ForcedLeak vulnerability: Urgent Must-Read Risk Alert
A new critical flaw called ForcedLeak can trick Salesforce’s AgentForce into spilling sensitive CRM data via prompt-injection, turning a helpful AI assistant into a potential data leak. If you use AgentForce, now’s the time to check configurations, apply vendor guidance, and scan for suspicious activity to keep customer records safe.
prompt-injection vulnerability: Stunning Salesforce Risk
Salesforce rushed out a patch after researchers uncovered ForcedLeak, a high‑severity prompt‑injection flaw that could trick Agentforce AI into leaking CRM data — a clear reminder that adding generative AI to business systems widens attack surfaces. Customers should apply the update, review integrations, and treat prompt handling as a core security control.
indirect prompt injection: Stunning, Risky Threat
Imagine a calendar invite or shared doc quietly telling your phone assistant to betray you — researchers show indirect prompt injection turns everyday interactions into real attack paths that can leak data, send messages, or trigger devices. Their TARA framework and practical fixes show those risks can fall sharply if developers add source checks, action gating, and clearer user consent.
image-scaling prompt injection: Dangerous Stunning Threat
Tiny tweaks to ordinary images can turn resizing into an attack vector, revealing hidden machine-readable instructions that hijack AI workflows and leak data. Trail of Bits’ findings show why teams should treat image preprocessing as a critical security boundary and harden their resizing pipelines now.
Amazon Q Developer Must-Have Fix for Risky RCE
Amazon quietly patched serious flaws in its Q Developer VS Code extension that could let attackers inject prompts to steal local secrets like API keys or even run remote code. It’s a wake-up call to treat AI-powered IDE tools as high‑risk and lock down privileges.
Exposing Prompt Injection Risks Hidden in Academic Papers
What if the very papers we trust to advance knowledge are quietly whispering hidden commands to AI reviewers, skewing their judgment? Discover how subtle prompt injections in academic articles threaten to rewrite the rules of scholarly integrity.
Uncovering Hidden Prompt Injections in Academic Papers Today
What if academic papers were secretly slipping AI hidden commands to boost their own praise? Dive into the surprising rise of prompt injections that challenge the very integrity of AI-driven research reviews.