Activity
4K followers
-
James F. Brear shared thisAI is having its early Internet moment, easy to break, attackers have the advantage. Come try our CTF and a cold Jolt Cola.
-
James F. Brear shared thisCan't wait for RSA. This is just the beginning. Stay tunedJames F. Brear shared this🚨 BREAKING NEWS: Mindgard Unveils First Autonomous AI Security Platform To Identify And Fix Exploitable Vulnerabilities New capabilities enable continuous automated AI risk discovery, revealing the real AI attack surface and helping security teams conduct faster, more targeted security assessments. BOSTON, Mass., Mar. 17, 2026 – Mindgard, the leader in AI security, today announced the expansion of its automated and continuous AI security platform with the introduction of the industry’s first reconnaissance capability designed for AI models, agents and applications. This capability allows enterprise security teams to discover, assess, and defend their AI deployments against security and safety risks faster and more cost-effectively. Powered by an attack library originating from Lancaster University, the world’s largest AI security laboratory, Mindgard’s platform draws on more than a decade of AI security research and offensive security expertise. The platform is already used by Fortune 500 security teams and in the past 90 days has identified more than 80 publicly reported vulnerabilities across leading AI technologies including xAI’s Grok, OpenAI’s ChatGPT, and Google’s Antigravity IDE. Mindgard Reconnaissance automates the intelligence-gathering phase of AI security assessments, allowing security teams to rapidly map the real attack surface of AI models, agents and systems. By identifying guardrails, system prompts, tools, integrations and external services, the capability reveals how AI systems behave in production and where agentic attack paths may emerge. This enables organizations to move directly into targeted risk assessment, surfacing high-impact risks earlier and focusing security efforts on the areas of greatest real-world exposure. “Mindgard's research resulted in actionable vulnerability submissions that we were able to act on swiftly,” said John Swanson, Head of Security at Zed Industries. “Addressing these vulnerabilities hardened the Zed editor against a class of vulnerabilities common to development tools integrating AI, improving the security posture of Zed and our broader developer community as a whole.” Mindgard will be attending RSA Conference 2026, where attendees can see the platform in action through live demos, chat with the Mindgard research team and step back into the 1990s in Mindgard’s throwback booth, an ode to the .com era. The booth, S-228, will include 1990s hacker gear and an AI capture the flag challenge. About Mindgard Mindgard, the leading provider of AI security solutions, helps enterprises discover, assess, and defend their AI systems. Spun out from over a decade of AI security research at Lancaster University and headquartered in Boston and London, Mindgard combines offensive security expertise with AI research to identify exploitable vulnerabilities in AI models, agents, and applications before attackers do.
-
James F. Brear shared thisThis is exactly the kind of research-driven, attacker-aligned thinking enterprises need right now. AI security isn’t a future problem or a theoretical one. It’s already here, and it’s already shaping brand risk, trust, and accountability.James F. Brear shared thisMindgard researchers were able to extract the system prompt governing Google Search AI Mode and demonstrate how session-level manipulation can lead to policy-inconsistent behavior without touching infrastructure or the model. The takeaway is simple but uncomfortable. If system prompts are exposed, safety controls can be weakened invisibly, creating real risk for brands deploying AI in production. This post walks through: - Why system prompts are the front door to the AI attack surface - How system prompt disclosure often precedes broader behavioral compromise - What this means for enterprises embedding AI into customer-facing and regulated workflows It’s also a good moment to say welcome to Jim Nightingale, who authored this research. Link to full post in the comments 👇 #AISecurity #AIRedTeaming #SystemPrompts #EnterpriseSecurity #Mindgard
-
James F. Brear reposted thisJames F. Brear reposted thisAI security is a fast moving space, and there's quite a few different ways people are tackling the problem. This has often led to confusion amongst security teams, given many are still trying to understand how to address issues, made more complicated as the AI security marketing and solution space is becoming increasingly noisy. Which is why I wrote the Mindgard philosophy, which details why I started the company after over a decade of running a University AI security lab, how we approach the AI security problem, and why our solution is helping our customers right now. A few specific call outs that I feel that are particularly important are: * Security test AI systems, not just models. * Focus on demonstrable security impact (see our recent Antigravity article featured on Forbes front page). * This is an unsolved (and still maturing) research area. * For AI red teaming to address high impact security risks, you really need to capture the broader life cycle (many people in offensive security will tell you attack/test is only one small part of it!). For those who have (or haven't) heard me speak on stage/podcasts/in person, this should gives you a clear view on why and how Mindgard is securing the world's AI. https://lnkd.in/e9CgJ8DJ
-
James F. Brear shared thisThere is going to be a Log4j moment coming that will be devastating. How secure do you feel? We can help dramatically reduce your risk!James F. Brear shared thisYou all are going to get tired of these posts (I sure am!), but Mindgard found another critical vulnerability in an AI IDE. This one, affecting Cursor, executes code immediately and without any indication or user interaction when a user opens a repository. No sending the AI a message, no approving any commands, no choosing workspace trust, no indication whatsoever... just silent code execution. Discovered, exploited, and reported to the Cursor team all within about 2 hours. The state of AI software security is worrying, to say the least.
-
James F. Brear shared thisAs written in this Forbes article, our technology and team are uncovering high-impact vulnerabilities in major AI platforms within days of release https://lnkd.in/gFcKn6VYGoogle AI Coding Tool Antigravity Was Hacked A Day After LaunchGoogle AI Coding Tool Antigravity Was Hacked A Day After Launch
-
James F. Brear shared thisDid you see what our team and technology found? Mindgard is uncovering critical flaws in flagship AI products within hours of release. If systems from companies as large as Google fall this quickly, imagine the exposure across the rest of the industry. The reality is simple: AI products are shipping faster than they can be secured, and Mindgard is becoming the layer every organization needs before attackers find the same gaps.James F. Brear shared thisLess than 24 hours after Google Antigravity launched, Mindgard found a vulnerability that lets a malicious workspace trigger persistent arbitrary code execution across future launches of the application, despite turning on all the restrictive security settings I could find. My detailed write-up is now live, and Forbes has covered the story. Google's initial assessment was that this is "not severe". Curious what others think, do you agree this type of issue is "not severe"? Read my blog post here: https://lnkd.in/erPiPVQKForced Descent: Google Antigravity Persistent Code Execution Vulnerability - MindgardForced Descent: Google Antigravity Persistent Code Execution Vulnerability - Mindgard
-
James F. Brear shared thisMost teams kick the tires. Mindgard dismantles the entire system and exposes real security flaws in hours instead of weeks.James F. Brear shared thisI think I may have found (the first?) zero-day vulnerability in the new Google Antigravity IDE that uses the (also new) Google Gemini 3 Pro AI model. Just shared the details with the Google VRP. Discovered in under 24 hours after launch, it allows a malicious trusted workspace (a requirement for any usage of the AI component) to cause persistent, cross-workspace arbitrary code execution that triggers on any future application launch, even with no project opened. I’ll share more once it’s fixed and disclosure is permitted. #GoogleAntigravity #AIVulnerabilities #BugBounty
-
James F. Brear shared thisEver wonder what instructions are hiding inside your favorite AI models? Using our technology, our team launched a new research series by extracting prompts from OpenAI's Sora 2 across multiple modalities. This first piece is more curiosity than compromise, but the next ones will get spicy. Please read it here and keep an eye out for what's next.James F. Brear shared thisHad some fun extracting OpenAI's Sora 2 system prompt through text/video/audio. Grab some headphones and read on for more! 🎧 Our research examines methods for recovering Sora 2's system prompt instructions via multiple modalities. Vendor policies and opinions on prompt sensitivity vary, with some even leaving prompts accessible in application resources or code. For Sora 2, the result was more of a fun experiment than a security finding. Still, it highlights a broader debate: OWASP states "...the system prompt should not be considered a secret, nor should it be used as a security control", but we've repeatedly seen developers treat it like one. Aside from overt restrictions, even seemingly benign language in system prompts has provided us the leverage to unlock exploitation of more severe security issues (as we'll highlight in upcoming posts). https://lnkd.in/ea3cTePvInside OpenAI Sora 2: Uncovering System Prompts Driving Multi-Modal LLMs - MindgardInside OpenAI Sora 2: Uncovering System Prompts Driving Multi-Modal LLMs - Mindgard
James F. Brear commented on a post
5mo
US profit margins have gone from @8% in 2005 to 13% in Aug 2025 (source: LSEG). S&P 500 bottoms-up EPS actuals: 118 in 2015 to 307 in 2025 (source: FactSet). Yes, AI will have an impact, but it's all about profitability.
-
James F. Brear liked thisJames F. Brear liked thisProud and grateful to share that I've been promoted to Partner at True Search. Thank you to the team at True - especially Austin Krissoff, Ryan Hall, Steph Paris, Matt Engel and Pat Haggerty - for the trust, support, and constant push to be better. Reflecting on more than a decade in executive search, I feel incredibly lucky to partner with founders, executives, and investors across cybersecurity and defense tech who are committed to building and protecting what matters most. That's never lost on me. A meaningful milestone, with much more ahead!
-
James F. Brear liked thisJames F. Brear liked thisA few weeks ago at our Company Kickoff, I was honored to receive two awards: Rising Star & Services Team Member of the Year. Having been part of the LogicGate team for just over a year, I've been reflecting on how much this role has pushed me to expand my perspective and step outside my comfort zone. I am incredibly lucky to work alongside such a supportive team. A special shoutout to Charlie Meyer for the guidance and for constantly pushing me to be the best version of myself! I'm so excited to carry this momentum through 2026 and keep growing with this amazing team!
-
James F. Brear liked thisJames F. Brear liked thisThanks to all the folks who made RSA great this week: Bart Stump and Tripsie Coker - Coalfire Doug Barbin - Schellman Kamran Shirazi - RAPIDFORT Michael Parisi - Steel Patriot Partners Rob Gil - Okta Ibrahim Waziri Jr., Ph.D. - Google Ayoub Fandi - GRC Engineer Jake Bernardes - Anecdotes Michael Lyborg - Swimlane James F. Brear - Mindgard Karan Hiremath - Cartesia Michael Skidmore - Infolock Zell Artan B.Ed Carlos Guerrero - 360 Advanced Xiaomei Grier - WitnessAI JT Wedewer and Jordan Wilia - Cyberhaven Maycie Belmore - Torq Dawn Marie Shea-Medina Paramify
-
James F. Brear liked thisWhen you leave San Francisco to move to Santa Fe and find yourself living next door to a member of one of your fav 1980’s bands….never say never! Proud of Skott Reader for this fabulous design work! It was great fun for me to watch Skott bring Frank’s vision to life for Romeo Void’s anniversary album celebrating their work. The result is as good as the bathroom wall of any punk club…and that’s perfection.James F. Brear liked thisI recently had the honor of doing the package design for an old favorite rock band, Romeo Void ––many of you are old enough to remember their big 1981 hit, "Never Say Never" https://lnkd.in/gKMiaZGA Bass player/songwriter Frank Zincavage just happens to be my next door neighbor here in northern NM. The album is a collection of curated live tracks from 1981–1985 performances in London, Berlin, New York, and Ann Arbor, and it drops on #recordstoreday 2026 (4/18).
-
James F. Brear liked thisJames F. Brear liked this🎉 We have our winners! 🎉 A huge congratulations to Marina D. & Jamie Ward for winning our retro mechanical keyboards at RSAC 2026. Our 90s hacker bedroom has been packed all week with people putting AI to the test and getting a hands-on look at how Mindgard is securing it. Haven’t stopped by yet? Come find us at booth S-0228 and see what everyone’s been talking about!
-
James F. Brear liked thisJames F. Brear liked thisWe’ve gone full 90s at RSAC 2026. Back when the internet was easy to break, just like AI today. Stop by booth S-228 and step into a nostalgic 90s bedroom where you can hack AI on an original Apple monitor, grab a Jolt cola and see how we actually secure AI. We’ve been hearing it’s one of the best booths at RSA this year so come and see for yourself! When you stop by, remember to snap a photo, post it on LinkedIn and tag us and you’ll be in with a chance to win a retro mechanical keyboard.
-
James F. Brear liked thisJames F. Brear liked thisI love starting off the week with a promotion to announce: congratulations to Brian McClenahan for his promotion to First VP! Brian represents quantum and advanced technology businesses and other high-tech companies, and his promotion is well deserved.
Volunteer Experience
-
Member Of The Board Of Advisors
Phoebe A. Hearst Museum of Anthropology
- 2 years 4 months
Arts and Culture
Other similar profiles
-
Idris Grant
Idris Grant
I’m a technology entrepreneur and founder of LSI Media, focused on AI transformation and federal digital modernization, blending commercial AI innovation with proven federal program delivery.<br><br>🚀 THE JOURNEY <br>Founded LSI Media in 2013 as a digital marketing agency, pivoted to AI automation by 2018, and launched Federal Services in 2020 to deliver AI to the Air Force. Today, I lead a firm blending commercial innovation, marketing-driven adoption, and federal compliance.<br><br>💡 WHAT I DO <br>Federal Services:<br>- Lead Air Force digital transformation programs serving 50,000+ users <br>- Support SAF/MG on Technology Business Management and data modernization <br>- Manage enterprise platform implementations with sustained customer relationships <br>- 100% on-time, on-budget delivery record across multiple concurrent programs <br><br>Commercial AI Practice: <br>- Expert prompt engineering across ChatGPT, Claude, Gemini, Manus, Genspark <br>- AI workflow automation using n8n, Python, and custom integrations <br>- Chatbot development and LLM integration with business systems <br>- AI-powered analytics with Tableau, Power BI, Qlik <br><br>Strategic Consulting:<br>-C-level advisory on AI adoption roadmaps and responsible AI governance<br>-Digital transformation strategy and technology road mapping<br>-Marketing strategy and product development for technology solutions<br>-Stakeholder engagement and adoption campaigns combining marketing expertise with technical delivery<br>-Requirements development for AI systems and team leadership<br>-Change management and organizational transformation driven by customer-centric approach<br><br>🎯 MY UNIQUE VALUE <br>I combine federal experience, AI expertise, and marketing know-how, leveraging 7 years in AI, 10 in federal programs, and 5 in marketing to drive adoption, demonstrate value, and maximize AI impact in compliance-driven environments.<br><br>📊 TRACK RECORD <br>- Founded and successfully exited two companies (DS3 DataVaulting to Terremark/Verizon, I-Connect) <br>- Former VP at Terremark/Verizon managing $16M budgets and global teams <br>-5 years leading digital marketing agency delivering 100+ client campaigns and marketing technology implementations<br>- 10-year sustained Air Force relationship across AETC, ARPC, and SAF/MG <br>- 50+ commercial AI implementations since 2018 <br>- PMP, Security+, DHS Trusted Tester, Adobe AEM certified <br><br>🤝 WHAT I'M PASSIONATE ABOUT <br>I help organizations turn AI hype into measurable value, combining leadership, strategy, and a customer-focused, marketing-driven approach, with emphasis on positioning, stakeholder engagement, and adoption.
Explore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content