Australia now has access to Anthropic’s Claude Mythos. It may improve cyber safety – but not for everyone

The Conversation · 🇦🇺 public Australia · Jun 04, 2026 · 841 words · By Andrew Cullen

Digital Divide/Resource Inequality AI Cybersecurity Risks National Security

headphones Listen to the eFinder podcast briefing

Ready to play

Daily briefing

What to know about Digital Divide/Resource Inequality

The article discusses Anthropic's limited release of 'Claude Mythos,' an AI tool designed to identify software vulnerabilities, to select organizations including the Australian government. It weighs the potential benefits for cybersecurity defenders against the risks of AI-driven attacks and the growing resource gap between large and small companies.

Propaganda risk 30%

Claims checked 10

Techniques found 3

Topics 3

Coverage spectrum

Coverage gap: Low Left coverage

Left0%

Center86%

Right14%

7 sources compared across this story cluster. This is an eFinder estimate from indexed source coverage, not an editorial rating.

What happened

Artificial intelligence (AI) giant Anthropic has expanded access to a highly advanced model deemed too dangerous for public release, including Australia in the select handful of users.

Why it matters

The large language model, known as Claude Mythos, is now being rolled out to an additional 150 organisations across 15 countries, including the Australian government and several local businesses, as part of Project Glasswing.

Common ground

In an era where large-scale AI launches are happening on a day-by-day basis, this limited, gradual release may seem particularly surprising.

Perspective signals

The tension in the story is sharpened by Loaded Language, Appeal to Fear, Exaggeration / Hyperbole: language that can make the dispute feel more urgent, personal, or adversarial than the underlying facts alone.

Follow-up questions

What new context would change how readers understand this Digital Divide/Resource Inequality story?
What evidence would most clearly confirm or weaken the claim that They did so by tricking AI chatbots into changing account details?
How does this story connect Digital Divide/Resource Inequality with AI Cybersecurity Risks over the next few days?

open_in_new Read the original article: https://theconversation.com/australia-now-has-access-to-anthropics-claude-mythos…

analyticsAnalysis

30%

Propaganda Score

confidence: 90%

Minor concerns. Some persuasive language detected, but largely factual.

psychologyPropaganda Techniques Detected

eFinder identified 3 propaganda techniques in this article. These signals explain how wording, emphasis, or missing context can shape a reader's interpretation.

warning

Loaded Language 90% confidence

Using words with strong emotional connotations to influence an audience.

Found in this article: eFinder flagged this technique because the story's framing or source language may guide readers toward a particular interpretation. Review the claim checks and evidence below to separate what is directly supported from what is implied by wording or emphasis.

Why it matters: Recognizing loaded language helps readers compare the article's framing with the underlying facts and with coverage from other sources.

warning

Appeal to Fear 80% confidence

Building support by instilling anxiety or panic in the audience.

Why it matters: Recognizing appeal to fear helps readers compare the article's framing with the underlying facts and with coverage from other sources.

warning

Exaggeration / Hyperbole 70% confidence

Overstating facts or claims to create a stronger emotional response.

Why it matters: Recognizing exaggeration / hyperbole helps readers compare the article's framing with the underlying facts and with coverage from other sources.

fact_checkClaims Checked

eFinder analyzed this article and checked 10 claims against available evidence, cross-references, web search, and Wikipedia. Here is what the fact-checking layer found.

check_circle Corroborated 6

info Single Source 2

help Insufficient Evidence 2

check_circle

Claim 1: “They did so by tricking AI chatbots into changing account details.”

CORROBORATED

Sources explicitly state that hackers gained access by tricking the AI support bot into swapping or changing the email address associated with the accounts.

travel_explore

web search NEUTRAL — In common usage, the term most often refers to someone who gains unauthorised access to computer systems, but it is also used for security researchers, skilled programmers, computer enthusiasts, and m…
https://en.wikipedia.org/wiki/Hacker

travel_explore

web search NEUTRAL — Hackers is a 1995 American crime thriller film directed by Iain Softley and starring Jonny Lee Miller, Angelina Jolie, Fisher Stevens, and Lorraine Bracco, with Jesse Bradford, Matthew Lillard, Lauren…
https://en.wikipedia.org/wiki/Hackers_(film)

travel_explore

web search NEUTRAL — Sep 15, 1995 · Hackers: Directed by Iain Softley. With Jonny Lee Miller, Angelina Jolie, Jesse Bradford, Matthew Lillard. Teenage hackers discover a criminal conspiracy with plans to use a computer vi…
https://www.imdb.com/title/tt0113243/

check_circle

Claim 2: “Of the 23,000 vulnerabilities flagged by Mythos, only 6,200 were estimated as high-risk by Mythos.”

CORROBORATED

Two independent web sources explicitly state that Mythos flagged 23,019 potential vulnerabilities, with 6,202 estimated as high or critical.

menu_book

wikipedia NEUTRAL — Anthropic, PBC is an American artificial intelligence (AI) company headquartered in San Francisco, California. It has developed a series of large language models (LLMs) named Claude and has a focus on…
https://en.wikipedia.org/wiki/Anthropic

menu_book

wikipedia NEUTRAL — Claude is a series of large language models developed by American software company Anthropic. Named after Claude Shannon, Claude was released as an AI-based chatbot in March 2023. It is also used in A…
https://en.wikipedia.org/wiki/Claude_(AI)

menu_book

wikipedia NEUTRAL — Claude Mythos is a large language model developed by Anthropic. Anthropic has not officially released the model to the public, citing safety and misuse concerns over its ability to find software vulne…
https://en.wikipedia.org/wiki/Claude_Mythos

+ 3 more evidence sources

info

Claim 3: “human experts could only validate two in every three of these vulnerabilities as high-risk.”

SINGLE SOURCE

While sources confirm the model found thousands of vulnerabilities, none of the provided evidence mentions the specific 'two in every three' validation rate by human experts.

travel_explore

web search NEUTRAL — Apr 13, 2026 ... We conducted cyber evaluations of Anthropic's Claude Mythos Preview and found continued improvement in capture-the-flag (CTF) challenges and ...
https://www.aisi.gov.uk/blog/our-evaluation-of-claude-mythos…

travel_explore

web search NEUTRAL — Apr 13, 2026 ... The model discovered thousands of high-severity vulnerabilities spanning every major operating system and web browser [1]. Over 99% of those ...
https://labs.cloudsecurityalliance.org/research/ai-vuln-disc…

travel_explore

web search NEUTRAL — Apr 13, 2026 ... We've established that combining LLM-driven security which uncovers a wide range of unknown vulnerabilities with an already unmanageable backlog ...
https://checkmarx.com/blog/checkmarx-application-security-gu…

help

Claim 4: “These emails tricked the AI into leaking large amounts of private and confidential information, without the email ever needing to be opened by a human.”

INSUFFICIENT EVIDENCE

No evidence was found in the provided search results regarding the mechanics of the alleged 'Echoleak' attack.

help

Claim 5: “A similar attack known as Echoleak last year revealed how tying Microsoft Copilot to email accounts could introduce significant risks.”

INSUFFICIENT EVIDENCE

No evidence was found in the provided search results regarding an attack called 'Echoleak' or Microsoft Copilot email risks.

check_circle

Claim 6: “hackers used Meta’s AI powered chatbot to gain access to high-profile Instagram accounts, including Barack Obama’s.”

CORROBORATED

Multiple independent sources report that hackers used Meta's AI chatbot to gain access to high-profile Instagram accounts, specifically mentioning Barack Obama's White House page.

menu_book

wikipedia NEUTRAL — Facebook is an American social networking service owned by the American technology conglomerate Meta Platforms. It was founded in 2004 by Mark Zuckerberg, along with his Harvard College roommates and …
https://en.wikipedia.org/wiki/Facebook

menu_book

wikipedia NEUTRAL — Mark Elliot Zuckerberg (; born May 14, 1984) is an American businessman and programmer who co-founded the social media service Facebook and its parent company Meta Platforms. He is its chairman, chief…
https://en.wikipedia.org/wiki/Mark_Zuckerberg

menu_book

wikipedia NEUTRAL — Marne Lynn Levine (born 1970) is an American businesswoman. She was the chief business officer at Meta Platforms (formerly Facebook, Inc.) until February 2023. Previously, she was the first chief oper…
https://en.wikipedia.org/wiki/Marne_Levine

+ 3 more evidence sources

check_circle

Claim 7: “the Australian Signals Directorate welcomed Australia’s inclusion in Anthropic’s Project Glasswing.”

CORROBORATED

Multiple sources confirm that the Australian Signals Directorate (ASD) welcomed Australia's inclusion in Project Glasswing.

menu_book

wikipedia NEUTRAL — The Australian Cyber Security Centre (ACSC), the successor to the Cyber Security Operations Centre, is the Australian Government's lead agency for cyber security. The ACSC is part of the Australian Si…
https://en.wikipedia.org/wiki/Australian_Cyber_Security_Cent…

menu_book

wikipedia NEUTRAL — The Australian Signals Directorate (ASD), formerly the Defence Signals Directorate, is a statutory agency of the Government of Australia responsible for signals intelligence, providing intelligence su…
https://en.wikipedia.org/wiki/Australian_Signals_Directorate

menu_book

wikipedia NEUTRAL — The Special Operations Command (SOCOMD) is an Australian Defence Force command that was established on 5 May 2003 to unite all of the Australian Army's special forces units and by 2008 was fully opera…
https://en.wikipedia.org/wiki/Special_Operations_Command_(Au…

+ 3 more evidence sources

info

Claim 8: “Australians have repeatedly been the victims of costly cybersecurity incidents, including Optus, Medibank Private, the Melbourne International Film Festival, and Canvas.”

SINGLE SOURCE

The provided evidence confirms Optus is a major Australian company, but does not provide specific details or a list of cybersecurity incidents for Medibank Private, the Melbourne International Film Festival, or Canvas to corroborate the claim as a group.

menu_book

wikipedia NEUTRAL — Clare Ellen O'Neil (born 12 September 1980) is an Australian politician who is the Minister for Housing and Minister for Homelessness since July 2024, Minister for Cities since May 2025 and was the Mi…
https://en.wikipedia.org/wiki/Clare_O'Neil

menu_book

wikipedia NEUTRAL — This list of privatizations provides links to notable and/or major privatizations.
https://en.wikipedia.org/wiki/List_of_privatizations_by_coun…

menu_book

wikipedia NEUTRAL — Vocus Group, formerly known as Vocus Communications, is a telecommunications company headquartered in Melbourne, Australia. Vocus Communications was founded by James Spenceley as a wholesale, enterpr…
https://en.wikipedia.org/wiki/Vocus_Group

+ 3 more evidence sources

check_circle

Claim 9: “The large language model, known as Claude Mythos, is now being rolled out to an additional 150 organisations across 15 countries, including the Australian government and several local businesses, as part of Project Glasswing.”

CORROBORATED

Multiple independent web sources confirm that Anthropic is expanding Project Glasswing and the rollout of Claude Mythos to 150 organizations across 15 countries, specifically mentioning the Australian government.

menu_book

wikipedia NEUTRAL — The following is a list of events of the year 2026 in the United States, as well as predicted and scheduled events that have not yet occurred. July 4, 2026 was the 250th anniversary of the signing of …
https://en.wikipedia.org/wiki/2026_in_the_United_States

menu_book

+ 3 more evidence sources

check_circle

Claim 10: “Mythos is... an automated tool for assessing software to find critical bugs and vulnerabilities.”

CORROBORATED

Multiple sources describe Claude Mythos as a tool designed to find and fix software vulnerabilities and exploits autonomously.

travel_explore

web search NEUTRAL — Apr 7, 2026 ... Identifying vulnerabilities and exploits with Claude Mythos Preview ; SWE-bench Pro · 77.8% · 53.4% ; Terminal-Bench 2.0 · 82.0% · 65.4% ; SWE-bench ...
https://www.anthropic.com/glasswing

travel_explore

web search NEUTRAL — The goal is to use Claude Mythos Preview to find and fix vulnerabilities in critical software before adversaries develop comparable tools. Partners use the ...
https://www.contrastsecurity.com/glossary/mythos-ai

travel_explore

web search NEUTRAL — May 7, 2026 ... Claude Mythos is Anthropic's generative AI model designed for extended autonomous reasoning—and it's the first frontier model to demonstrate ...
https://www.endorlabs.com/learn/what-is-mythos-and-why-it-ma…

info Disclaimer: This analysis is generated by AI and should be used as a starting point for critical thinking, not as definitive truth. Claims are verified against publicly available sources. Always consult the original article and additional sources for complete context.