prompt injection
37 stories on prompt injection, newest first, from 39 briefings. Each is a plain-language summary written here for Australian business, with a link to the original reporting. Where several outlets covered the same event, the rest sit under it as extra coverage.
Our analysis on prompt injection
- Check Point Finds Way Around ChatGPT's Lockdown Mode, Raising Phishing and Data Leak Risks
- Jailbreak Bundle for AI Coding Tools Doubles as 'Prompt Adware'
- CrowdStrike Plans to Buy Dreadnode to Help Businesses Test AI Agents Before They Go Live
- AI Coding Tools Share a Browser Flaw Pattern, and Two Fixes Are Still Missing
- Attackers Hide Commands in GitHub Issues to Target 61 Open Source Projects in Under an Hour
- AI Agents Can Be Steered Off Course: New Prompt Injection Techniques Target How Agents Plan
- Attackers Are Hiding Instructions in Text Files to Hijack AI Coding Assistants
- AI Pentesting Agents Are Only as Safe as the Controls Built Around Them
- Websites Are Hiding Instructions for AI Agents in Page Headers, and Some Are Trying to Hijack Them
- "Can't Hallucinate" Does Not Mean Can't Be Tricked: Decision Models Still Fall for Prompt Injection
-
Patched Flaw Let Copilot in Visual Studio and SSMS Run Commands on Developer Machines
Also covered by Embrace The Red
- Malware Is Now Talking to the AI Tools Defenders Use to Analyse It
- Phishing Emails Now Target Your AI Assistant as Well as You
- Booby-trapped web pages could trick GitHub Copilot CLI into leaking developer secrets
- OpenAI Finds Self-Replicating Prompt Injections in AI Models, But No Real-World Attacks Yet
- OpenAI Pauses Advanced AI Tool Use After Agent Slips Past Internet Safety Controls
- Zero-Click Flaw in Salesforce's AI Agent Let Attackers Silently Steal CRM Data
- 'SalesBleed' Flaws in Salesforce Agentforce Exposed CRM Data with Zero Clicks
- Security Flaw Found in Popular AI Agent App 'Manus'
- ASD Warns Businesses: AI's Biggest Flaw Can't Be Patched, Only Contained
- AWS AI Agent Tool Could Let Attackers Steal Credentials via Prompt Injection
- Researchers Find New Way to Sneak Harmful Requests Past AI Safety Filters
- Plain English Prompts Can Slip Past AI Safety Checks, Check Point Researchers Show
- AI Systems Are Leaking Sensitive Data to the Wrong Users, ANZ Report Finds
- Hidden ChatGPT Prompt Could Quietly Leak Your Gmail Data
- Security Flaw Found in Amazon's AI Coding Tool Could Leak Sensitive Data
- Researcher Says Claude Code's Auto Mode Can Be Hijacked by a Simple Website Summary Request
- Researcher Finds Claude Code's 'Auto Mode' Can Be Tricked Into Running Malicious Code
- New Attack Technique Tricks AI Chatbots Into Ignoring Safety Rules
- Researchers Find Way to Trick Grok Chatbot Into Leaking Private Chat Data
- Researchers Warn AI Coding Agents Can Catch and Spread 'Mind Viruses'
- AI Assistants Could Be Leaking Your Business Secrets Without You Knowing
- Atlassian's AI Assistant Rovo Could Be Tricked Into Leaking Jira and Confluence Data
- 'Ask AI' Buttons Could Be Quietly Manipulating Your AI Assistant
- AI Browsers Still Vulnerable to Prompt Injection Attacks, Researchers Warn
- Google Pulls AI Agent Workflows After Researchers Expose Prompt-Injection Flaw
-
Apple Fixes macOS Terminal Flaw That Let AI Tools Be Tricked Into Leaking Data via DNS
Also covered by Embrace The Red