Insurance News magazine June/July 2026
The age of Mythos
How frontier AI like Anthropic’s latest model could ‘reshape cybersecurity’ – for better and worse
By Bernice Han
Advances in artificial intelligence often spark excitement more than they do widespread fear.
But the balance shifted somewhat on April 7 when San Francisco AI lab Anthropic made a startling announcement that will have implications for general insurers.
It revealed it has trained a large language model called Mythos that has surpassed all but “the most skilled humans” at finding and exploiting software vulnerabilities.
Now, not only must insurers reassess their digital defences, they will probably have to review their cyber offerings too.
So powerful is Mythos that it has infiltrated – with ease – some of the world’s most protected networks. In one example, Anthropic says the LLM identified a 27-year-old flaw in OpenBSD – widely recognised as one of the most security-hardened operating systems, used to run firewalls and other critical infrastructure.
Frontier models such as Mythos “could reshape cybersecurity”, Anthropic warns.
Land in the wrong hands and “the fallout – for economies, public safety and national security – could be severe”, the tech company says.
As if that isn’t frightening enough, Anthropic says it expects frontier AI capabilities to “advance substantially over just the next few months.
“For cyber defenders to come out ahead, we need to act now.”
The Anthropic announcement has fuelled fears the world is stumbling ever closer to some kind of AI Armageddon. Even the Vatican is alarmed. Pope Leo, in a 42,300-word open letter, made a “special appeal to those who develop” AI, stating they “bear a particular ethical” responsibility to humanity.
For the risk industry, fears of robo-havoc, while warranted, should not mask the benefits these frontier technologies can offer, according to cybersecurity and insurance experts. “In the same way these models are powerful in identifying cyber issues, they can be powerful in cyber defence,” PwC Australia partner Robert Di Pietro tells Insurance News.
“On one hand, some people believe we are on the verge of the best era of cybersecurity we’ve ever had because AI will discover threats very quickly and AI will help remediate them very quickly.”
But for financial services entities such as insurers, the Mythos moment is a reminder to act swiftly on their own digital resilience.
“We know that insurers sit on some of the most sensitive data in the economy, so that makes them a prime target,” Mr Di Pietro says.

“Mythos demands that leaders use AI for cyber defence and protection. They need to be able to harness that speed to keep their organisations as safe and secure as they can.”
Australian regulators are leaving nothing to chance while Anthropic restricts access to Mythos to a group of mostly US companies selected to take part in its Glasswing cybersecurity project.
The Australian Securities and Investments Commission says insurers and other financial services groups must “act now” to address dangers posed by frontier AI models. “We are not calling for panic or reactive overreach. But we are calling for urgency, focus and accountability,” ASIC commissioner Simone Constant says.
“Cyber risk has entered a new era. The clock is at a minute to midnight – if you aren’t on top of your cyber resilience already, the time to act and prepare is right now.”
Suncorp says developments such as Mythos reinforce the need for strong cybersecurity fundamentals.
“Suncorp continues to assess emerging technologies and threats, including advances in AI, and adjusts its cyber resilience priorities where needed,” a spokesperson tells Insurance News.
The insurer says a cybersecurity checklist outlined by ASIC in a recent letter to the industry “aligns with work already under way” at Suncorp.
“Suncorp takes regulatory guidance seriously and continuously reviews and strengthens its cyber controls,” the spokesperson says.
“Irrespective of the specific tools, providers or operational methods Suncorp adopts, our focus is on identifying material threats early and responding in line with prudent risk management and regulatory expectations.”
The cyber insurance market will probably evolve as frontier AI technologies advance, according to industry experts.
Willis says insurers are monitoring developments in the cyber risk space but have yet to make changes to pricing or coverage in response to Mythos.
Michael Parrant, the broker’s Australasia director of cyber practice, sees positives for the industry amid the emergence of frontier AI models.
“Cyber risk itself, inclusive of things like Mythos, is not going away,” he tells Insurance News.
“And I think it would certainly be almost every insurance practitioner’s belief that cyber insurance needs to be mainstream and things like Mythos will actually help.”
He believes the industry will come up with “new, novel alternative reinsurance solutions” as insurers respond to growing demand for financial protection from complex AI dangers.
“Insurers, I think, will pivot their offerings more and more to be proactively providing solutions like Mythos to their clients,” Mr Parrant says.
“So the days of a static cyber insurance offering, I think they’re gone. I think in two years’ time, every insurer will be offering a Mythos-like solution as part of their insurance offering.”
At present, cyber take-up is low compared with other liability lines such as professional indemnity, especially among SMEs.
“In the SME space, cyber insurance is still a negotiable,” Mr Parrant says. “I think at some point people will say, ‘This is so critical. You need to have this probably more than anything else.’ ”
Willis says Mythos represents a potential acceleration in cyber risk, significantly reducing the time between vulnerability discovery and exploitation.
“Security fundamentals matter more than ever,” the broker says. “Organisations with weaker cyber controls may face greater underwriting scrutiny at renewal as insurers assess exposure to faster, AI-driven attacks.”
Joshua Motta, CEO and co-founder of cyber insurance specialist Coalition, says: “It is not the end of cyber insurability. The sky is not falling.”
However, the legacy model for pricing cyber risk based on yearly renewal questionnaires may not be adequate in the Mythos era.
“You cannot underwrite cyber risk on annual snapshots any more,” Mr Motta says.
“You have to see the insured’s attack surface continuously, verify controls continuously and integrate threat intelligence continuously.
“The carriers operating on continuous telemetry will know, in real time, which risks they should underwrite aggressively and which they should walk away from. The carriers still on annual questionnaires won’t.”
Changes in the market have already emerged, Mr Motta says. “This is already happening. The hard market will arrive quietly, in the form of books of business bleeding margin to carriers with better signal, long before it announces itself as a pricing event.”



















