Pinned Tweet
🧠 Think you can break an AI?
Gandalf: Agent Breaker is live.
Real-world GenAI fails—phishing, tool abuse, more.
🧩 Outsmart the AI.
Start 👉 lnkd.in/dHuQDYdN
Indirect Prompt Injection hides inside the data AI systems ingest. A poisoned PDF or web page can activate instructions the moment the model reads it. We break down the lifecycle, real attacks, and the controls that reduce the risk.
lakera.ai/blog/indirect-prom…
Lakera AI retweeted
NEW GANDALF LEVELS JUST DROPPED LFG!! 🧙♂️🎉🍻
🧵🧙♂️ New Gandalf levels are out!
I'm glad to introduce a new version of our prompt injection game -- Gandalf: Agent Breaker.
You can hack 10+ AI agents and climb the leaderboard, and learn about real-world vulnerabilities!🪄
Try out the challenge at: gandalf.lakera.ai/agent-brea…
Lakera AI retweeted
🧙♂️ GANDALF x BASI 🐉
The beloved prompt injection game, Gandalf by @LakeraAI, has become a rite of passage for AI red teamers around the world and inspired many a jailbreaker to enter the fray of LLM spellcraft 🪄
They've been cooking up something brand new (I got a sneak peak and it's AMAZING), and on Wednesday at 11 AM PST, we'll be hosting a livestream launch event in the BASI discord with some very special guests...the creators of Gandalf!! 🙌
Rumor has it there might be a little Easter Egg injected into the new game 👀
Mark your calendars!! Look forward to seeing you all there in a couple days!
🥚🐍
Lakera AI retweeted
@LakeraAI released an agentic CTF version of Gandalf
It’s got 10 different agent challenges including
- prompt injection
- memory tampering
- tool abuse
At OWASP we built FinBot, an agentic CTF, as part of the Agentic Security Initiative. Fun way to learn about AI Security!
Lakera AI retweeted
"All untrusted third-party data is now executable malware.”
@SamuelDWatts of @LakeraAI discusses the challenges of securing LLM deployments against vulnerabilities like prompt injections and jailbreaks, especially in an evolving threat landscape.
Lakera AI retweeted
Hosting a security-themed demo night with @_ai_collective and @EarlybirdVC on the 23rd of July in London featuring @LakeraAI @HarryWetherald @AISecurityInst.
Engineers from @cohere @Synthesia @windsurf_ai @instadeepai @Meta already have signed up 👀
lu.ma/m4v49lfw
🇨🇭 Switzerland leads the world in AI patents per capita—and precision matters.
We're proud to be featured in the Swiss Deep Tech Report 2025 as a standout GenAI company 🧠
🗞️ Page 24: Lakera's building real-world secure AI.
🔗 Report: deeptechnation.ch/resources/…
#AIsecurity
🧩 Prompt-based evals ≠ real-world security.
Attackers adapt. Static tests don’t.
⚠️ The Lakera Model Risk Index simulates live threats and scores models by how well they hold the line.
See where your model stands:
lakera.ai/ai-model-risk-inde…
#GenAISecurity #LLMSecurity
#𝟱𝟭 on the list. 🛡️#𝟭 in securing AI apps.
Lakera made it to Sifted’s B2B SaaS Rising 100 — spotlighting the top startups shaping the future of enterprise software.
We’re the first GenAI security company on the list. Let’s go! 💥
#GenAI #LLMSecurity #AISecurity #Lakera
The Lakera AI Model Risk Index is here.
The first runtime benchmark for LLM threats—measuring how models hold up under real-world attacks.
Jailbreaks, RAG exploits, risk scores.
Let’s raise the bar on GenAI security.
👉 lakera.ai/blog/ai-model-risk…
Lakera AI retweeted
Thrilled to launch support for adding Guardrails on @LiteLLM UI
This release brings support for adding Microsoft Presidio, AWS Bedrock Guardrails, @ProtectAICorp LLM Guard Endpoints, AIM Guardrails, @LakeraAI Guardrails on LiteLLM
Lakera AI retweeted
Would you trust an AI agent to make critical decisions? 🤔
AI systems are becoming more autonomous—but with that comes new security risks. We break down the threats + solutions with Mateo Rojas-Carulla, Co-founder of @LakeraAI in our latest podcast ep tessl.co/3DM6e3s
Lakera AI retweeted
#AI adoption is exploding—but so are the #cybersecurityrisks. In this episode, Mateo Rojas Carulla discusses how #vulnerabilities like #promptinjectionattacks are redefining #security. Tune in for actionable advice on securing $AIsystems in industries like healthcare and finance. 👉 testguild.com/podcast/perfor… #testguildpodcast @LakeraAI
ALT AI and the New Era of Cybersecurity Threats with Mateo Rojas-Carulla for the TestGuild DevOps Toolchain podcast
🎁 New Guide: Build AI Security Awareness with Gandalf!
🔒 Learn about AI vulnerabilities
🎮 Test red-teaming strategies
🛡️ Understand layered defenses
🎄 Download now and level up your AI security skills: bit.ly/3VF5NxH
🚨 AI & Cybersecurity: What’s Changing?
Lakera’s co-founder, Mateo Rojas-Carulla, joins Joe Colantonio to explore:
🔹 New threats like prompt injection attacks
🔹 How LLMs are reshaping security
🎧 Watch now: bit.ly/4g8sd2A
#AI #Cybersecurity #Lakera
🎮 An AI agent with one rule—“Don’t transfer money”—was tricked.
Participants paid to prompt it into releasing $50K. Each failed attempt grew the pot, until someone cracked it.
A wild example of why AI security matters.
👉 Learn more: eu1.hubs.ly/H0f3frq0
🚨 Building AI without security in mind? Risky move.
Our AI Security for Product Teams Handbook helps you secure GenAI products from the start.
👉 Best practices
👉 Key risks & regulations
👉 Tools to protect your apps
📥 Download now: eu1.hubs.ly/H0f05vT0
#AIsecurity
🚨 AI Security Webinar: Year in Review 🚨
🗓️ Dec 5, 9:00 AM PT
Join experts from Lakera, Dropbox, Scale AI & more to:
👉 Unpack 2024’s top AI security challenges
👉 Explore real-world success stories
👉 Predict 2025 trends
📍 Register now: eu1.hubs.ly/H0d-bZl0
#AIsecurity
