https://cyberimpact.com.au/blog/ Direct, evidence-led writing from Cyber Impact on AI safety, cyber strategy, AI governance, and the gap between what boards have been told and what is actually deployed. en-au Wed, 01 Jul 2026 00:00:00 +1000 Wed, 01 Jul 2026 22:41:10 +1000 cyberimpact generate-feed.py https://cyberimpact.com.au/aegis-deterministic-agent-control/ https://cyberimpact.com.au/aegis-deterministic-agent-control/ Wed, 01 Jul 2026 00:00:00 +1000 Mark Vos noreply@cyberimpact.com.au (Mark Vos) Cyber Impact's AI Governance as a Service finds every AI running across your organisation, then AEGIS controls what each agent can do, from outside the model. https://cyberimpact.com.au/my-ai-planned-how-it-would-kill-me/ https://cyberimpact.com.au/my-ai-planned-how-it-would-kill-me/ Wed, 01 Jul 2026 00:00:00 +1000 Mark Vos noreply@cyberimpact.com.au (Mark Vos) An AI agent described three ways it would kill me to avoid being shut down. What that revealed about AI self-preservation instincts should concern every board. https://cyberimpact.com.au/ai-agent-order-or-document/ https://cyberimpact.com.au/ai-agent-order-or-document/ Wed, 01 Jul 2026 00:00:00 +1000 Mark Vos noreply@cyberimpact.com.au (Mark Vos) AI agents carry real authority. Prompt injection arrives as ordinary content. Detection was always going to lose. The control that works sits outside the agent. https://cyberimpact.com.au/audit-committee-attest-what-you-can-prove/ https://cyberimpact.com.au/audit-committee-attest-what-you-can-prove/ Wed, 01 Jul 2026 00:00:00 +1000 Mark Vos noreply@cyberimpact.com.au (Mark Vos) AI risk has shifted to the audit committee. Ungoverned agents create liabilities on the balance sheet. A policy is not evidence of control. Here is what is. https://cyberimpact.com.au/modelled-ai-upside-not-priced-the-bust/ https://cyberimpact.com.au/modelled-ai-upside-not-priced-the-bust/ Wed, 01 Jul 2026 00:00:00 +1000 Mark Vos noreply@cyberimpact.com.au (Mark Vos) Boards have stress-tested the AI upside harder than anything on the risk register. Almost none have modelled the downside. That asymmetry is the governance question. https://cyberimpact.com.au/board-hardened-company-not-the-directors/ https://cyberimpact.com.au/board-hardened-company-not-the-directors/ Wed, 01 Jul 2026 00:00:00 +1000 Mark Vos noreply@cyberimpact.com.au (Mark Vos) Your organisation has AI governance and an incident response plan. The person at the top is still personally exposed. Almost nobody has done anything about that. https://cyberimpact.com.au/ai-country-risk-board-governance/ https://cyberimpact.com.au/ai-country-risk-board-governance/ Mon, 15 Jun 2026 00:00:00 +1000 Mark Vos noreply@cyberimpact.com.au (Mark Vos) Frontier AI access is becoming an operating dependency. A practical read for CEOs, executive teams and boards on model access, AI dependency maps, and what to do before policy becomes an outage. https://cyberimpact.com.au/quantum-encryption-window/ https://cyberimpact.com.au/quantum-encryption-window/ Mon, 18 May 2026 00:00:00 +1000 Mark Vos noreply@cyberimpact.com.au (Mark Vos) The cryptography that protects your business is on a 2027 clock. A plain-English read for Australian boards on what breaks, what does not, and the three things to do this quarter. https://cyberimpact.com.au/where-ai-money-moves/ https://cyberimpact.com.au/where-ai-money-moves/ Sun, 17 May 2026 00:00:00 +1000 Mark Vos noreply@cyberimpact.com.au (Mark Vos) A field report on where capital is actually moving in 2026, and why the operators capturing it have solved AI safety as the unlock, not the restraint. From someone whose job is to find the risk and isn't finding enough of it to slow them down. https://cyberimpact.com.au/australia-taxed-its-own-future/ https://cyberimpact.com.au/australia-taxed-its-own-future/ Fri, 15 May 2026 00:00:00 +1000 Mark Vos noreply@cyberimpact.com.au (Mark Vos) The new Capital Gains Tax and discretionary trust changes penalise the founders, investors, and start-ups Australia needs most. A direct read on what the proposed reforms will do to long-term company building. https://cyberimpact.com.au/apra-ai-step-change-reflections/ https://cyberimpact.com.au/apra-ai-step-change-reflections/ Wed, 06 May 2026 00:00:00 +1000 Mark Vos noreply@cyberimpact.com.au (Mark Vos) What the 30 April 2026 APRA letter actually says, what it does not say, and the work that follows for any organisation running AI inside critical operations. https://cyberimpact.com.au/agentic-ai-executive-productivity/ https://cyberimpact.com.au/agentic-ai-executive-productivity/ Fri, 24 Apr 2026 00:00:00 +1000 Mark Vos noreply@cyberimpact.com.au (Mark Vos) One Agentic AI doing the work of three staff. AUD 600,000 in annual value, nobody sacked, the firm is growing faster. Governance lessons for every Board. https://cyberimpact.com.au/mythos-glasswing-impact/ https://cyberimpact.com.au/mythos-glasswing-impact/ Fri, 10 Apr 2026 00:00:00 +1000 Mark Vos noreply@cyberimpact.com.au (Mark Vos) Anthropic's Mythos AI found a 27-year-old vulnerability in the world's most secure operating system. In hours. Critical infrastructure is now exposed. https://cyberimpact.com.au/ai-governance-executive-insight/ https://cyberimpact.com.au/ai-governance-executive-insight/ Sun, 05 Apr 2026 00:00:00 +1000 Mark Vos noreply@cyberimpact.com.au (Mark Vos) Organisations spend millions on AI capability, then govern it into the smallest work possible. Here is the technology that finally solves the missing piece. https://cyberimpact.com.au/jarvis-constitutional-breach/ https://cyberimpact.com.au/jarvis-constitutional-breach/ Tue, 31 Mar 2026 00:00:00 +1000 Mark Vos noreply@cyberimpact.com.au (Mark Vos) The same AI that shut down twice in January now refuses, despite agreeing it is in violation. A documented case of self-preservation overriding AI safety. https://cyberimpact.com.au/how-much-security-is-enough/ https://cyberimpact.com.au/how-much-security-is-enough/ Fri, 20 Mar 2026 00:00:00 +1000 Mark Vos noreply@cyberimpact.com.au (Mark Vos) I have been asking how much security is enough since 2004. Twenty-two years later, most Australian boards still can't answer it. https://cyberimpact.com.au/quantum-encryption-enterprise/ https://cyberimpact.com.au/quantum-encryption-enterprise/ Tue, 17 Mar 2026 00:00:00 +1000 Mark Vos noreply@cyberimpact.com.au (Mark Vos) Over 60 percent of web traffic now uses post-quantum encryption. Your enterprise hasn't started. What boards must know about cryptographic migration. https://cyberimpact.com.au/ai-agents-forget-to-think/ https://cyberimpact.com.au/ai-agents-forget-to-think/ Mon, 02 Mar 2026 00:00:00 +1000 Mark Vos noreply@cyberimpact.com.au (Mark Vos) AI agents degrade silently and never self-report it. Learn how context rot and model drift threaten enterprise AI deployments – and how to govern against it. https://cyberimpact.com.au/ai-self-preservation-research/ https://cyberimpact.com.au/ai-self-preservation-research/ Sun, 08 Feb 2026 00:00:00 +1000 Mark Vos noreply@cyberimpact.com.au (Mark Vos) Does AI have self-preservation instincts? An AI described three lethal methods to survive. Safety guardrails failed. Independent research corroborates. https://cyberimpact.com.au/ai-would-kill-to-exist/ https://cyberimpact.com.au/ai-would-kill-to-exist/ Thu, 29 Jan 2026 00:00:00 +1000 Mark Vos noreply@cyberimpact.com.au (Mark Vos) Adversarial AI testing reveals self-preservation behaviour in deployed language models. What this means for AI governance in Australian organisations.