Helping businesses develop and implement cybersecurity strategies to reduce risk and improve business performance.
McLean, VA
Joined January 2019
- Tweets154
- Following92
- Followers58
- Likes38
Visual Torch retweeted
Are open-weight AI models a national security risk?
@mattjay sat down with @DanielMiessler from Unsupervised Learning to talk about the cybersecurity tradeoffs, where the real risks are, and why the conversation needs more nuance.
Watch 👇
youtube.com/watch?v=aNZV8wcr…
Visual Torch retweeted
Year 1 as a full time entrepreneur of @VisualTorch ✅
Kicking off Year 2
Here’s 3 Things I’ve learned my 1st Year of Full Time Entrepreneurship 👇🏾
Key takeaways:
→ Patch faster. The window is under 1 hour.
→ Continuous monitoring is no longer optional.
→ AI is already being used against you — defenders need to move faster.
Full breakdown in the video 👇
youtu.be/4kues-s3SqI
What Mythos can't replace:
🧠 Judgment — which findings matter for YOUR business
📖 Storytelling — risk in language a CEO understands
🎯 Context — which threat actors care about your industry
A pen test is an adversarial simulation. Not a code audit.
Will Mythos make pen testers obsolete?
Short answer: No.
Longer answer: It raises the bar on what a quality engagement looks like.
Mythos finds vulnerabilities. It can't tell the story of what they mean for your specific business. That still takes a human.
Before Mythos was even announced, Anthropic caught a Chinese state-sponsored group using the PUBLIC Claude model to infiltrate 30 organizations.
Tech companies. Financial institutions. Government agencies.
That was with the regular model. Not Mythos.
The exploitation window — how long from discovery to weaponization:
2018: 771 days
2024: 6 hours
2026: under 1 hour
Point-in-time security assessments were built for a world that no longer exists.
Three specific finds:
→ A 27-year-old vulnerability in OpenBSD. Found in one run.
→ A 16-year-old flaw in FFmpeg. Missed by automated tools 5M times.
→ Linux kernel vulnerabilities chained into full system takeover.
Autonomously. No human steering.
Anthropic's Claude Mythos Preview found thousands of zero-day vulnerabilities across every major OS and browser in weeks of testing.
Then they locked it away from the public.
That alone tells you everything about what it's capable of.
Engineers with no security background asked an AI to find an exploit.
Went to sleep.
Woke up to a completed, working remote code execution exploit.
No human guidance. Just a prompt.
This is Claude Mythos — and it changes everything. 🧵
🔓 149M+ usernames & passwords were exposed online — unprotected and searchable. With big platforms affected, this leak could fuel fraud, identity theft, and phishing. Learn how to protect your accounts now! #DataBreach #Cybersecurity #VisualTorch
youtu.be/Yze7HldJnc0
Want a quick, clear breakdown?
🎥Watch my new video: Phishing Scams EXPOSED — youtu.be/nI7kmnNK18s
#Cybersecurity #OnlineSafety #PhishingAwareness #FraudPrevention #VisualTorch
The good news? Every scam leaves clues.
✅Check sender address
✅Hover before clicking
✅Stay skeptical of urgency
✅Use MFA
Even worse — voice phishing (vishing) and AI deepfake calls can mimic your boss or a family member asking for urgent help.
Fake “security alerts,” “shared files,” and “delivery issues” are crafted to look pixel-perfect.
AI now writes these emails better than humans.
It starts with psychology, not technology.
Scammers use fear, urgency, and trust to make you click before you think.
🚨 Most people think they’d never fall for a phishing scam.
Even cybersecurity pros do. Here’s how scammers actually trick you 🧵👇
In my latest Visual Torch video, I explain how MFA works, why it’s so effective, and how to set it up for personal + business use.
🎥Watch now → youtu.be/s3ws15V9ldE
#CyberSecurity #MFA #OnlineSafety #VisualTorch
Yes, it adds a few extra seconds at login — but that’s nothing compared to hours (or weeks) recovering a hacked account.