AI isn’t just about productivity tools and chatbots any more. Some of the most capable models are now being talked about in the same breath as critical infrastructure, national security and protecting systems at scale.
That’s where Anthropic’s Mythos has attracted attention, and controversy. So what’s going on, and does it matter for NZ businesses?
Why Mythos set off alarm bells
Recent coverage around the world, including reporting here in New Zealand, has highlighted concern about just how powerful Anthropic’s newest models have become, and how they’re being positioned.
The controversy isn’t about one feature. It’s about capability and intent. In broad terms, Mythos is described as:
- a framework that shapes how Anthropic’s most advanced models behave
- a way to constrain, direct and monitor AI used in high-impact settings
- a way to keep AI decision-making in line with human values and safety limits
That all sounds sensible. So why the worry?
The real tension
It comes down to one question: should AI be trusted to analyse, protect or influence critical infrastructure at scale?
Some commentators are uneasy because these models can reason across complex, interconnected systems, may find weaknesses humans haven’t spotted, and their recommendations could shape real decisions very quickly. The same capability that makes them useful makes them powerful, and power without visibility makes people nervous.
How it’s being positioned
From public statements and reporting, Anthropic’s intent appears to be defensive, not operational. Mythos-guided models are framed as tools to analyse large, complex systems for hidden risk, test assumptions people take for granted, and find systemic weaknesses before they’re exploited. Think of it as an extremely advanced audit brain, not something that takes over the controls.
What sort of weaknesses can AI find?
Public reporting doesn’t list specific vulnerabilities found by Mythos-driven models, and I won’t guess. But in general, AI-assisted analysis is very good at spotting:
- settings that have drifted across large environments
- “privilege creep” and accounts with far more access than they need
- hidden dependencies between systems that make a problem spread further
- single points of failure nobody owns
- old assumptions baked into long-standing processes
None of these are new problems. They’re just hard for people to see all at once.
What it means for everyday NZ businesses
Most businesses won’t be directly touched by Mythos. You won’t be installing it or managing it, and you won’t suddenly be under AI-driven controls. But you’ll feel the effects further down the line.
As AI shapes what vendors and platforms do, expect more frequent security hardening advice, better prioritisation of which vulnerabilities really matter, and updates that fix whole classes of problem rather than single bugs. You’ll see it in Microsoft security baselines, identity and access policies and cloud configuration standards.
A tsunami of updates?
No sudden tidal wave, but a steady increase. More frequent incremental security updates, tighter default settings over time, and more “recommended” changes becoming “required”. A gradual tightening, not a flood.
The real takeaway: stay proactive
The Mythos debate isn’t really about Anthropic. It’s about a bigger shift: AI is getting better at seeing risk before people can. For businesses, that reinforces a familiar message:
- patch regularly
- review who has access to what
- keep your systems on supported versions
- don’t ignore the “minor” warnings
Staying proactive keeps you in step with where the industry is heading, without any drama.
My view
It’s healthy to question powerful technology. But don’t confuse capability with control. Mythos is an attempt to put guardrails around advanced AI, not take them away. The controversy shows how new this territory is, not that businesses are about to lose control of their systems.
For NZ businesses the focus stays the same: strong basics, trusted platforms, clear accountability and good local advice. That’s how you stay secure, however clever the tools get.
Want to make sure your basics are in good shape? Book a free health check, or call us on 0800 688 768.