Five Eyes Alliance Declares AI Is the Best Way to Combat Increasingly Dangerous AI

Following a meeting of its cybersecurity heads, the usually reticent Five Eyes intelligence sharing alliance released a 22 June 2026 call to action regarding the “evolving landscape of artificial intelligence”, as while AI is creating defensive tools at an accelerating pace, these advances are at the same time increasing the potential for “malicious actors” to weaponise AI technologies against nation-states.

The Five Eyes, a 1946-established pact between the United States, the United Kingdom, Canada, New Zealand and Australia, has somewhat conveniently found the solution to these AI woes via the use of artificial intelligence to deliberately strengthen defence and “not just improve efficiency”. Yet, this goal may be easier for US cyber nerds to advance than their counterparts down under.

The reason for this is that US AI company Anthropic removed access to Mythos 5, the most powerful cybersecurity tool anywhere, to all partners outside of the US, following an order from the Trump White House on 12 June 2026, which was the date of its release. And this occurred, just a week after the Australian government and a few companies had received access to the AI model’s predecessor.

The order for Anthropic to remove access to Mythos 5 globally, came just days after US president Donald Trump signed a 2 June 2026 executive order, requiring AI companies to hand over their frontier AI models for US government analysis prior to release, and this marked a significant shift from that administration’s policy to allow AI companies to develop their technologies unfettered.

So, as the Trump administration has accepted that “no holds barred” AI might pose it a national security threat, Australia only appears able to get its foot in the door partway in terms of the pending AI revolution and its security benefits, and even with the Five Eyes call to action, Australian reliance on US government-controlled AI platforms will ultimately leave it vulnerable to such withdrawals.

Five Eyes cybersecurity

“Frontier Al models are anticipated to exceed current industry expectations, fundamentally transforming both offensive and defensive cyber capabilities,” explained the heads of cybersecurity representing the nations making up the Five Eyes alliance last week. “The timeline is not years, it is months.”

“AI is not a future consideration – it is already here,” the experts continued and went on to explain that the technology “lowers barriers for malicious actors and increases the speed and complexity of attacks, shrinking the window between vulnerability discovery and exploitation ever more quickly”, whilst “at the same time, AI offers powerful tools to strengthen defence”.

The Five Eyes release further posits that AI can no longer be confined to just a “tech issue”, as the risks it poses are so grave that “boards and executives” should have responsibility over it, and as these “systems evolve, new and previously unknown vulnerabilities will emerge”, and being prepared will help those in control to “prevent escalation into major operational and financial crises”.

Practical actions the Five Eyes experts recommend include not connecting systems to AI that don’t need to be, accelerating the time it takes to patch vulnerabilities, that old technological systems be addressed as they remain an easy to target, while critical systems should constantly be reviewed and strengthened and those overseeing AI systems should be in constant preparation for incidents.

“Cyber resilience is not an IT issue,” the cybersecurity heads said in ending their statement, “it is central to operational continuity and market trust. Leaders who act now will reduce exposure, strengthen resilience, and build confidence with customers, partners and investors. Those who delay will face growing and avoidable risk.”

The artificial intelligence quagmire

The Australian government released its National AI Plan in December 2025, which aims to grow a local AI industry and establish an AI-enabled economy. In line with this vision, Anthropic’s top lawyer Jeffrey Bleich came to Australia in May to meet with Home Affairs and the Australian Signals Directorate, with the aim of incorporating Mythos locally, which was a goal achieved last month.

Australia is incorporating AI technologies into the protection of its designated systems of national significance (SoNS), which is a small subset of critical infrastructure assets that have been determined as the most crucial entities in the nation, due to their interdependence across sectors. And laws passed in 2018 permit the home affairs minister to privately declare entities as SoNS.

The April 2026 released Claude Mythos Preview AI model was incorporated into certain Australian systems in early June. However, this model was superseded by the release of Mythos 5 the following week, and that was then taken off the market immediately. Mythos 5 was accompanied by Fable 5, a restricted AI model for commercial use, and it was made available over here on 1 July.

The reason the original Mythos was not initially available in Australia and other countries on release was it was considered “too dangerous”.

Anthropic CEO Dario Amodei has this month penned an essay about the dangers of AI and the need for regulation, in light of Trump’s executive order that determined to accelerate the industry but also to have the government monitor it for new vulnerabilities.

In his report, Amodei explains that while in 2024 there were concerns that AI could eventually, produce biological weapons or be capable of “autonomous misbehaviour”, those risks are now “clearly here”. And he considers that AI should be treated and tested like aeronautical technology, and the most powerful systems ought to be approached in a similar way to nuclear power.

The AI CEO further considers that as the technology becomes “a more general economic substitute for human cognitive abilities”, there could be disruption to labour markets, and “enduring job displacement” should be minimised, as such dramatic shifts to employment in general could be countered via data tracking, pro-employment incentives and perhaps, a universal basic income.

Another key aspect identified as producing problematic incidents involve the safeguard mechanisms currently in place for older technologies, such as drug research, as these could slow down developments produced by AI, so more “radical and flexible” ways to facilitate such advances in technologies should be contemplated.

The final issues that Amodei identifies in relation to the AI revolution is the risk it poses in terms of placing too much power in the hands of the state, which may undermine civil liberties. And an example Amodei identified in terms of AI eroding basic rights was in the area of “fully autonomous weapons” and in contemplating their use both in military situations and domestically.

But the CEO considers that such problems could be avoided if democratic nations bandy together to prevent problematic AI developments.

The irony of Amodei delivering these statements whilst living under an increasingly authoritarian Trump administration is likely lost only upon a few.

The AI bonanza down under

Australian communities are currently becoming increasingly aware of the threat the AI revolution poses in terms of the environmental impact as proposals for datacentres in various locations are increasingly being considered.

One of the largest datacentres in the Asia Pacific is set to be constructed on Ngadjuri land in South Australia’s Bundey, and locals fear its impacts, especially in terms of the water usage involved in the running of the plant.

Another proposal for a datacentre was to be established on Gundungurra and Dharug land in a residential area of the New South Wales town of Katoomba, but it was recently averted due to public outcry.

Australia already has over 160 datacentres, mainly located in NSW and Victoria, and there are at least another 90 more facilities being considered.

Australian federal police commissioner Krissy Barrett recently flagged utilising AI chatbots for law enforcement purposes online as she’s concerned about terror networks and organised crime using artificial intelligence and echo chamber algorithms to locate and then radicalise local children of around 13 to 16 years old, with the ultimate aim of getting them to commit criminal offences.

Barrett’s solution is to employ AI chatbots to track down these nefarious actors, and she further told the Murdoch press that she’s been in consultation with one tech company about the potential for such a system to be set in motion, and she was also planning to raise the potential for AI to fight crime using AI chatbots at a Five Eyes law enforcement meeting taking place a fortnight ago.

So, it appears that just like the Five Eyes heads of cybersecurity have determined, those in charge of law enforcement are coming to the same conclusion, which is the best way to prevent the dangers that artificial intelligence poses humanity is to employ artificial intelligence against them.

Although the benefits the AI revolution is promising do tend to appear to favour the billionaire class, which is increasingly consolidating its power globally, and this bias seems especially stark when considering that jobs may soon be scare, and the majority will then be left to hold out for the good will of those at the top to commence providing all with a universal basic income.

Main image: Shutterstock 

Image ID 2793303583