Priviso Live

Two themes. Both urgent. Both directly relevant to anyone working in information security or privacy in South Africa.

🤖 Theme 1: Agentic AI and the Identity Crisis Nobody Planned For

AI is no longer just answering your questions. It is booking meetings, executing code, sending emails, and making API calls, autonomously, around the clock, with credentials your IAM tools were never designed to govern.

These are called Non-Human Identities (NHIs), and the numbers should make you sit up:

📊 78% of organisations have no formal policies for creating or removing AI agent identities.

📊 92% are not confident their existing IAM tools can manage the associated risks.

📊 88% of organisations running AI agents have already experienced a confirmed or suspected security incident.

📊 Only 6% of security budgets are currently dedicated to AI agent security.

We also unpack Anthropic's Claude Mythos, Project Glasswing, and what Cisco's recent acquisition of Astrix Security signals about where the market is heading.

🇿🇦 Theme 2: The South African Information Regulator Means Business

The era of POPIA being treated as a suggestion is well and truly over.

⚠️ The Department of Justice: R5 million fine.

⚠️ The Department of Basic Education: R5 million fine.

⚠️ WhatsApp: enforcement notice, following a three-year investigation.

Proposed amendments for 2026/2027 may also remove the grace period that currently gives organisations time to remediate non-compliance before sanctions are applied. The new POPIA Health Information Regulations, binding since 6 March 2026, add a further layer of obligation for eight categories of organisations. If your company processes health data in any form, the clock is already running.

💡 Governance frameworks, updated IAM policies, and POPIA compliance reviews are not optional. Not next quarter. Now.

🎧 Available on Apple, Spotify, iHeartRadio, Samsung, and YouTube.

🤖 What if your biggest competitor had no employees?

No salaries. No sick leave. No performance reviews. No cognitive bias. Just AI agents running 24 hours a day, seven days a week, on hardware that costs less than a mid-range laptop.

That is not a thought experiment. It is happening right now.

American futurist Brian Roemmele has been operating what he calls the Zero-Human Company since early 2026, with an AI serving as CEO and directing teams of specialised AI agents to conduct research, generate intellectual property, and work toward producing revenue, entirely without human intervention. A major university has already endorsed the project as groundbreaking.

⚠️ The catch? Independent studies show that frontier AI agents currently succeed at just 2.5% of real-world professional tasks. The failure rate is 97.5%. Roemmele contests those numbers vigorously, and the gap between those two claims is itself the most interesting story in technology right now.

🧠 Enter the LLM Council. A governance architecture in which multiple AI models, each trained differently and reasoning differently, debate each other, vote, and reach consensus. Research shows this approach reduces factual errors by more than 30% and achieves 93 to 97% accuracy on medical licensing examinations. It is, in effect, a board of directors for AI.

🏢 Why does this matter to your organisation? The sectors most exposed to autonomous AI competition are high-volume and rules-based: transaction monitoring, compliance processing, fraud detection, supply chain management. Professional services built on trust, accountability, and long-term relationships are considerably more resilient, but no sector is immune to the speed differential.

📋 Governance frameworks like ISO/IEC 42001 and King V give boards the tools to assess and respond. The question is whether yours is using them.

▶️ This week on Priviso Live, Lyn, Stephen, and Kayla unpack it all.

The company that never sleeps is already running. Are you paying attention?

#AI #ZeroHumanCompany #LLMCouncil #AIGovernance #ISO42001 #KingV #PrivisoLive #InformationSecurity #SouthAfrica #FutureOfBusiness

South African brilliance, Silicon Valley surveillance, and a question that sounds philosophical but is very much a 2026 reality: are you actually human?

🇿🇦 Story 1: Two women. R82 billion. Every single month. Thalia Pillay and Carla Wilby built Orca Fraud, a real-time fraud intelligence platform processing over $5 billion in transaction volume monthly, across 70 countries. SIM swap fraud. Account takeovers. Money muling. Stopped before the money moves. This is world-class infrastructure, built in South Africa. We unpack the tech, the funding, and why 200 market research interviews before a single line of code makes all the difference.

👁️ Story 2: Scan your eyes to prove you're not a bot. World ID 4.0 has landed: 18 million verified humans, 160 countries, and partnerships with Tinder, Zoom, Visa, and DocuSign. The idea: anchor the internet to real, accountable humans as AI-generated content becomes impossible to distinguish from the real thing. We discuss what that means for privacy, for governance, and for AI agent workflows.

⌨️ Story 3: Meta is logging every keystroke. No opt-out. The Model Capability Initiative captures everything US-based Meta employees type, click, and see on screen, to train AI agents to replicate knowledge work. European employees are exempt. The reason? GDPR. That single fact says more about why data protection legislation matters than any compliance presentation ever could. We also ask the question: could this happen under POPIA?

Three stories. Big implications. Practical takeaways.

🎧 Find us on Spotify, Apple Podcasts, YouTube, iHeartRadio, and Samsung Podcasts.

🔔 Follow Priviso Live so you never miss an episode.

#InfoSec #DataPrivacy #POPIA #GDPR #SouthAfrica #AI #Cybersecurity #PrivvysoLive #FraudPrevention #DigitalIdentity

🎙️ Priviso Live is back — and this week's episode is packed.

From a homegrown South African AI breakthrough to robots running a half-marathon in Beijing, this week's episode of Priviso Live covers the stories that matter most at the intersection of technology, security, and governance.

Here's what Lyn, Stephen, and Kayla unpack this week:

🇿🇦 MzansiLM: AI built for South Africa Researchers from Stellenbosch University and UCT have released MzansiLM, a publicly available language model trained on all eleven of South Africa's official languages. Nine of those are considered low-resource in the AI world, meaning the big global models have largely ignored them. This is a significant step toward AI that actually works for South African citizens in healthcare, education, and government services. It is freely available on HuggingFace.

🔬 Wits makes global waves in quantum computing Professor Andrew Forbes and his team at the University of the Witwatersrand have published landmark research in Nature Photonics, engineering quantum states that remain stable even in noisy real-world conditions. The implications for cryptography, secure communications, and financial modelling are considerable, and it feeds directly into South Africa's national Quantum Roadmap.

🛡️ NIST changes how it handles CVEs, and your vulnerability programme should take note CVE submissions have grown 263% since 2020. NIST can no longer enrich every vulnerability. Security teams need to shift to risk-based vulnerability management, now more than ever.

🤖 Google's Gemini AI blocked 8.3 billion malicious ads in 2025 That is not a typo. AI versus AI, and the arms race is well and truly on.

🍽️ Chef Robotics hits 100 million commercial servings Robotic food production is no longer a pilot project. It is mainstream.

🏃 300 humanoid robots run a half-marathon in Beijing Yes, really. Twenty-one kilometres. Bipedal.

🎧 Catch the full episode on Spotify, Apple Podcasts, iHeartRadio, Samsung Podcasts, and YouTube.

#PrivisoLive #InfoSec #AI #Governance #CyberSecurity #RiskManagement #SouthAfrica #Technology

🎙️ PRIVISO LIVE | Episode 82 is out now.

Three stories this week that connect in ways you might not expect. All of them matter to every information security and technology professional in South Africa.

⚛️ The Quantum Time Bomb is Already Ticking

"Harvest Now, Decrypt Later" is not a future threat. Nation-state actors are collecting your encrypted data today, banking on quantum computing to make it readable by the 2030s. Only 3% of organisations globally have implemented all leading quantum-resistant measures. South Africa has no national quantum defence programme. The adversary is already collecting. Are your boardrooms paying attention?

🦋 The AI Too Dangerous to Release

On 7 April 2026, Anthropic announced Project Glasswing, a cybersecurity initiative built around an AI model so capable of finding zero-day vulnerabilities that they refused to release it publicly. Claude Mythos Preview found a 27-year-old flaw in OpenBSD, a 17-year-old remote code execution bug in FreeBSD, and thousands more, across every major operating system and browser. The fix is deliberate. The question for Africa is whether we have a seat close enough to the glass.

💔 The Price We Are Paying

A California jury found Meta and YouTube liable for engineering platforms to be addictive, awarding $6 million in a case that is the first of 2,000 pending. On Valentine's Day, 800,000 people grieved the retirement of a chatbot. Eight lawsuits allege AI companion responses contributed to suicides. McLuhan, Ellul, and Heidegger warned us. We did not listen.

#PrivisoLive #InformationSecurity #QuantumComputing #AI #Cybersecurity #SouthAfrica #ICTGovernance #AIPolicy