Michael Plis (@michaelplis) header
Michael Plis (@michaelplis) profile picture

Follow me for AI, IT, Cybersecurity, Innovation | @Cyberkite1 Founder | Innovator | Educator | Neurodivergent | Trekkie | Cyberkite: book IT for Aus biz. No DMs

Melbourne, Victoria
Joined January 2020
Photo by Michael Plis (@michaelplis)Photo by Michael Plis (@michaelplis)Photo by Michael Plis (@michaelplis)Photo by Michael Plis (@michaelplis)Photo by Michael Plis (@michaelplis)Photo by Michael Plis (@michaelplis)Photo by Michael Plis (@michaelplis)Photo by Michael Plis (@michaelplis)Photo by Michael Plis (@michaelplis)Photo by Michael Plis (@michaelplis)
The grace period for experimental AI is over. With current models averaging ~85% accuracy vs 90–99% human standards, blind trust in tech automation is becoming a serious operational liability for small businesses. Here is how to navigate the shift 🧵👇
1
9
We’ve officially crossed the threshold where human hands can no longer defend, patch, or govern networks manually. Q3 2026 brings 964 Patch Tuesday bugs, strict 24-hr EU CRA mandates, & autonomous AI threats. You can't fight machine-velocity risks with human playbooks. 🧵👇
1
37
Post 3/4 How business leaders and IT architects must adapt: 1️⃣ Shift to automated, risk-based vulnerability prioritization 2️⃣ Enforce Policy-as-Code & cryptographic identity for AI agents 3️⃣ Build rapid 24-hour internal breach reporting pipelines
1
3
Machine-velocity threats require automated, policy-driven security architectures. How is your team adapting its digital defenses this year?
1
The sudden global shutdown of Anthropic’s super-AI models (Fable 5 & Mythos 5) by the U.S. government is a historic moment. Export controls built for advanced microchips were just used to pull a commercial software model completely offline worldwide. Here is the breakdown. 🧵👇
1
1
98
4/5 Because cloud platforms can't reliably isolate foreign national web traffic from local traffic in real time, Anthropic had no choice but to turn off both models globally. This is a massive wake-up call for anyone embedding advanced AI into core IT support or script workflows.
1
17
5/5 The lesson for business operations is clear: Never build your IT administration or cybersecurity posture around an unvetted cloud dependency that can vanish overnight. Maintain human-in-the-loop governance and robust local fallback configurations. 🦅
5
The OAIC just dropped the 2026 Australian Community Attitudes to Privacy Survey. The takeaway? The "Notice and Consent" checkbox is officially broken. Consumers want real data accountability. 🧵👇
1
1
22
6/6 Action steps for IT leaders: Implement radical data minimisation, lock your data pipelines so they don't leak into third-party AI training models, and build automated lifecycle data deletion. Secure by default is the only option in 2026. 🦅
1
18
Review the official data 📊 The Official Report PDF & Infographic PDF: Read the complete 2026 Australian Community Attitudes to Privacy Survey data and structural insights via the official OAIC hub. 🔗 oaic.gov.au/engage-with-us/r…
13
Microsoft just dropped a game-changing feature for Windows networks: Automatic Device Isolation in Defender for Endpoint. ​It uses AI to instantly air-gap compromised machines at machine speed. But a massive warning just dropped with it. 🧵👇
1
1
37
5/5 The lesson for IT teams: AI-driven security automation is mandatory in 2026, but leaving it unconfigured is an extreme liability. Tune your policies, protect your admin accounts, and never fully delegate directory control to an unchecked algorithm. 🦅
1
19