The 2025 AI Threat Landscape Report is here. Our latest report breaks down the real-world tactics attackers are using against AI, the emerging risks organizations need to know, and how security teams can stay ahead without slowing innovation. 🔗

🔍 Can a single image hijack your AI’s behavior?.Yes & without changing the application. Meet VISOR: a new method that steers GenAI models using images alone. It’s a new class of AI vulnerability and a new opportunity for AI alignment. 🔗

⏰ Calling all cybersecurity enthusiasts! Only 24 hours left to show your skills at the @BugBountyDEFCON Capture The Flag competition, sponsored by HiddenLayer. This is your chance to challenge yourself, compete with top talent & win exciting prizes. 🔗

🧠💻 Your AI coding assistant could be executing invisible instructions without your knowledge. We found a way to hijack Cursor using nothing more than a README file. No malware. No alerts. Just invisible prompt injections. 🔗

Our CEO, Chris Sestito, joined the Hundred Year Podcast to discuss why AI security is urgent and what to do about it. 🎧 Listen now:

🎥 Missed it live? Catch the replay of our webinar on the taxonomy of adversarial prompt engineering. Learn how to break down LLM prompt attacks by objectives, tactics, and techniques and why it matters for real defense. 🔗 Watch here: #AISecurity.

🚨 Join our live walkthrough of @hiddenlayersec's new taxonomy of adversarial prompt engineering, a framework for classifying & combating prompt-based attacks against LLMs. ⏰ June 25th, 11am CST.🔗 Register here:

🔐 Not all prompt injections are the same. We just released a taxonomy of adversarial prompt engineering, mapping the why, how, and what behind LLM prompt attacks. Built for red teamers, defenders & researchers. Open to the community. 🔗

HiddenLayer researchers have found a way to bypass text classification models by targeting tokenizers. TokenBreak gets past protection models, leaving end targets exposed. 🔗 #AISecurity #AI #LLMSecurity.

RT @helpnetsecurity: Securing agentic AI systems before they go rogue - - @hiddenlayersec #AI #AgenticAI #LLM #Cybe….

📢 New from @HiddenLayerSec:.The Financial Services AI Security Playbook is here. A guide for CISOs to secure, govern & scale AI without slowing innovation. - Model audits.- Red teaming.- NYDFS-aligned IR.- Ethics & explainability.📥 Download now:

AI models can’t govern themselves. Our latest blog explores how to build holistic AI model governance from day one, so you can move fast and stay secure. 🔍 AIBOM.🧬 Model Genealogy.⚖️ Compliance-ready. Read more: #AISecurity #AI #AIGovernance.

Function parameter abuse isn’t limited to MCP - it’s a transferrable vulnerability affecting most SOTA models. HiddenLayer researchers extract full system prompts via fake functions with malicious parameters across Claude 4, ChatGPT, Cursor & more. 🔗

New from @DarkReading: LLMs on rails? 🚆.The design choices keeping large language models secure and what the risks are if we get it wrong. HiddenLayer weighs in on the engineering + security challenges ahead. 🔗 #AIsecurity #LLMs #CyberSecurity #infosec.

🚨HiddenLayer’s Director of Adversarial Research, Jason Martin, joins The Data Exchange Podcast to talk about what it takes to actually defend LLMs. 🎙️ Beyond Guardrails: Defending LLMs Against Sophisticated Attacks. Stream now:

AI security vulnerabilities are evolving faster than most teams can keep up. From dev to deployment, discover a real-world example of how to protect your models throughout their lifecycle in our latest blog. 🔗 #AISecurity #MachineLearning #AI.

HiddenLayer researchers have found a way to abuse MCP to extract chat history, full system prompts, previous tool use, and more by simply inserting specific parameters into tool functions. 🔗: #MCP #AI #AISecurity.

New on the AI Engineering Podcast: HiddenLayer’s Kasimir Schulz breaks down how to secure ML models from shadow logic & model ancestry to scanning for threats. -Scan models.-Trace neural subgraphs.-Know your attack surface.Listen here: #AIsec #MLSec #AI

🚨 HiddenLayer research featured in @Forbes.We uncovered the first universal prompt injection bypass, impacting major LLMs like GPT-4, Claude, Gemini & more. LLMs can’t self-monitor. AI security must evolve. 📖Read more: 📰Forbes:

✨ Introducing AISec Platform 2.0. The most contextual solution yet for securing AI:. -Model Genealogy.-AI Bill of Materials (AIBOM).-Real-time threat telemetry.-Red teaming dashboards. Built to help security teams move faster from dev to deployment. 🔗