T.J. Marlin (@AIRiskExpert) header
T.J. Marlin (@AIRiskExpert) profile picture

AI Risk Expert | CEO @Guardrail_Tech | Empowering people to harness AI securely + ethically, while staying in control

Park City, UT
Joined August 2025
Photo by T.J. Marlin (@AIRiskExpert)Photo by T.J. Marlin (@AIRiskExpert)
And now it's the NSA’s classified network. Coincidence across this many critical systems in about a week feels statistically improbable, which points to something shared underneath. Assume the next failure is coming and test your fallback now. @Guardrail_Tech @BretBaier @brianstelter
12
A tech giant offline, flights grounded, banks disrupted. Outage after outage has been explained as routine, and none has been attributed to AI so far. I’m skeptical that settles it when most companies can’t see where AI operates inside their own walls. Know your AI and control it before it becomes a headline.
1
1
31
A digital war is underway as we speak, and it has been since at least 2019. The FBI and CISA confirmed that Chinese state actors broke into major American phone carriers through Salt Typhoon, while Volt Typhoon sat inside our power, water, pipeline and transportation systems for five years before investigators found it. North Korea has stolen billions in crypto to fund its weapons program. That's the backdrop for what happened this week. @Guardrail_Tech @BretBaier @brianstelter
1
53
The battlefield is concentration. A handful of providers run our communications, security, cloud and AI, and we're wiring autonomous agents into all of it at machine speed. In a real conflict, the side that can see, bound and stop its own automated systems keeps operating.
1
13
What a company can do today is know what it could still operate if its largest provider went dark, record every consequential AI action, test the dependencies behind its backup plans and keep an independent way to stop automated behavior that's gone wrong. The companies that do will treat the next outage as an inconvenience instead of a headline.
5
Appreciate @FoxBusiness taking on this topic. Protecting executives from influence belongs in every security program for AI and national security leaders.
Replying to @FoxBusiness
T.J. Marlin, CEO of Guardrail Technologies, tells FOX Business that for executives in AI and industries tied to national security, "the exposure goes well beyond a stolen document. Insight into research priorities, customers or deployment plans tells an adversary where a company is heading, and sometimes where a country is heading. An influenced decision or a well-placed introduction carries strategic value on its own."
22
Anthropic’s warning today should end one debate: There is no kill switch for AI. AI is already inside finance, defense, healthcare, energy and critical infrastructure. Perimeter security isn’t enough. The next security layer must live inside AI—watching agents, detecting change and enabling real-time human intervention. @jesenhuangg @sundarpichai @sama @DarioAmodei @finkd
1
5
13,628
Joined @FINTECHTVglobal this morning on our new research into the AI cyber-disclosure gap. 97% of S&P 500 companies discuss AI. Only 1 in 5 document a process for managing its cyber risk. Fewer than 4 in 100 describe a governed one. The findings are stark.
1
1
27
A question we explore alongside boards and executives every day: we're promoting AI across the business, but can we actually stand behind what we're disclosing about its risk? @Guardrail_Tech published the first study to measure that gap across S&P 500 filings. The findings are stark. Report and Board Diagnostic linked.
1
1
26
Big day on the frontier. A US government export control directive tonight forced Anthropic to suspend Fable 5 and Mythos 5 for all customers, citing a security concern. Its other models kept running. The lesson: frontier AI access can change overnight. Know what your AI touches.
2
75
AI agents can be persuaded, not just hacked. What makes this particularly dangerous is that while being persuaded, they may still be operating within their approved limits. Without the right limits in place, everything is permitted. This is a problem that @Guardrail_Tech was built to solve. Read more in the @techradar article below. Meta AI's recent hack is a terrifying wake-up call for anyone who puts their trust in AI systems | TechRadar
1
39
Companies are deploying internal AI agents faster than they can govern them, and those agents are a new kind of insider threat. Employees across teams are building agents with no visibility into the data they expose or the access they open. I spoke to the WSJ about it - link below.
1
1
24
Turncoat AI Agents Emerge as the New Inside Hackers - WSJ
1
15