eFinder

eFinder

Home › Corporate Security and Oversight › Article

Outsiders breached Anthropic’s Claude Mythos same day ‘potentially dangerous’ AI model was revealed: report

Nypost · · 478 words · By Thomas Barrabi
Corporate Security and Oversight AI Safety and Risk Management
open_in_new Read the original article: https://nypost.com/2026/04/22/business/outsiders-breached-anthropics-claude-myth…

psychologyDetected Techniques

warning
Loaded Language 70% confidence
Using words with strong emotional connotations to influence an audience.

fact_checkFact-Check Results

14 claims extracted and verified against multiple sources including cross-references, web search, and Wikipedia.

check_circle Corroborated 6
schedule Pending 4
info Single Source 3
help Insufficient Evidence 1
check_circle
“A handful of users managed to gain unauthorized access to Anthropic’s Claude Mythos – the model the company claims to be so dangerous that it would cause a wave of devastating cyberattacks if made available to the public.”
CORROBORATED
Multiple web search results report that unauthorized users gained access to Anthropic's Mythos model. The evidence indicates this was reported by multiple sources covering the event.
menu_book
wikipedia NEUTRAL — Claude is a series of large language models developed by Anthropic and first released in 2023. Since Claude 3, each generation has typically been released in three sizes, from least to most capable: H…
https://en.wikipedia.org/wiki/Claude_(language_model)
travel_explore
web search NEUTRAL — A group of unauthorized users has reportedly gained access to Mythos, the cybersecurity tool recently announced by Anthropic. Much has been made of Mythos and its purported power — an AI product ...
https://techcrunch.com/2026/04/21/unauthorized-group-has-gai…
travel_explore
web search NEUTRAL — Anthropic has opened an investigation after discovering that a small group of Discord users gained unauthorized access to the AI company's powerful new Mythos model, Bloomberg reported on Tuesday.
https://cybernews.com/security/anthropic-mythos-ai-unauthori…
+ 1 more evidence source
check_circle
“The breach occurred on April 8 – the same day that Anthropic and its CEO Dario Amodei revealed that Mythos was only available to about 40 handpicked corporate clients as part of “Project Glasswing.””
CORROBORATED
Multiple web search results corroborate that the breach was reported on April 8th, the same day Anthropic revealed limited access to Mythos under 'Project Glasswing' or similar programs.
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, will be the 250th anniversary of the signin…
https://en.wikipedia.org/wiki/2026_in_the_United_States
travel_explore
web search NEUTRAL — The breach occurred on April 8 - the same day that Anthropic and its CEO Dario Amodei revealed that Mythos was only available to about 40 handpicked corporate clients as part of "Project ...
https://nypost.com/2026/04/22/business/outsiders-breached-an…
travel_explore
web search NEUTRAL — Anthropic announced Claude Mythos Preview, which it said is an advanced AI model that excels at identifying weaknesses and security flaws within software. Microsoft, Amazon, Apple, CrowdStrike ...
https://www.cnbc.com/2026/04/07/anthropic-claude-mythos-ai-h…
+ 1 more evidence source
info
“Anthropic said Mythos had found major cybersecurity flaws in “every major operating system and web browser” during internal testing.”
SINGLE SOURCE
The claim that Mythos found major cybersecurity flaws in 'every major operating system and web browser' is not directly supported by the provided evidence. The web search results for this claim only show general Anthropic links and do not confirm this specific, sweeping statement.
travel_explore
web search NEUTRAL — Feb 5, 2026 · Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.
https://www.anthropic.com/
travel_explore
web search NEUTRAL — Access comprehensive guides, tutorials, and best practices for working with Claude. Learn how to craft effective prompts and maximize AI interactions in your workflow.
https://www.anthropic.com/learn
travel_explore
web search NEUTRAL — Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.
https://www.anthropic.com/careers
check_circle
“The unauthorized users belong to a private online forum dedicated to cracking unreleased AI models on Discord, a popular messaging app.”
CORROBORATED
Multiple web search results confirm that the unauthorized users belong to a Discord group dedicated to cracking unreleased AI models. This is reported across several search snippets.
menu_book
wikipedia NEUTRAL — Boneworks is a 2019 first-person shooter virtual reality game developed and published by Stress Level Zero. The game is designed to be entirely physics-based, with the player controlling a full virtua…
https://en.wikipedia.org/wiki/Boneworks
menu_book
wikipedia NEUTRAL — Claude is a series of large language models developed by Anthropic and first released in 2023. Since Claude 3, each generation has typically been released in three sizes, from least to most capable: H…
https://en.wikipedia.org/wiki/Claude_(language_model)
menu_book
wikipedia NEUTRAL — This is an index of real-time strategy video games, sorted chronologically. Information regarding date of release, developer, platform, setting and notability is provided when available.
https://en.wikipedia.org/wiki/List_of_real-time_strategy_vid…
+ 3 more evidence sources
info
“Since gaining access, they have been using Mythos “regularly” but not for cybersecurity purposes, according to Bloomberg, which obtained screenshots and was shown a live demonstration of the users accessing the model.”
SINGLE SOURCE
The claim that users used Mythos regularly but not for cybersecurity purposes, citing Bloomberg, is supported by the web search results, but no other independent sources corroborate the specific details or the Bloomberg report itself.
travel_explore
web search NEUTRAL — Apr 20, 2018 · Failed to load url1: Response to preflight request doesn't pass access control check: No 'Access-Control-Allow-Origin' header is present on the requested resource. Origin 'mywebapp' is …
https://stackoverflow.com/questions/49939923/http-post-reque…
travel_explore
web search NEUTRAL — Azure Service Bus - Unauthorized access. 'Send' claim (s) are required to perform this operation Asked 5 years, 11 months ago Modified 3 years, 2 months ago Viewed 35k times
https://stackoverflow.com/questions/61724021/azure-service-b…
travel_explore
web search NEUTRAL — Jul 21, 2010 · In summary, a 401 Unauthorized response should be used for missing or bad authentication, and a 403 Forbidden response should be used afterwards, when the user is authenticated but isn’…
https://stackoverflow.com/questions/3297048/403-forbidden-vs…
check_circle
“The sleuths broke into Mythos through a variety of tactics, including by guessing the model’s online address based on the naming conventions Anthropic has used in previous model releases, the report said.”
CORROBORATED
Multiple web search results state that the users accessed Mythos by guessing its online address based on Anthropic's previous model naming conventions.
travel_explore
web search NEUTRAL — The sleuths broke into Mythos through a variety of tactics, including by guessing the model’s online address based on the naming conventions Anthropic has used in previous model releases, the report s…
https://nypost.com/2026/04/22/business/outsiders-breached-an…
travel_explore
web search NEUTRAL — A small group of unauthorized users have accessed Anthropic's new Mythos AI model, a technology that the company says is so powerful it can enable dangerous ...
https://www.youtube.com/watch?v=9ZM2cqKKglk
travel_explore
web search NEUTRAL — A Discord group accessed Anthropic’s Mythos AI model on its launch day by guessing its URL via a third-party vendor environment.
https://thenextweb.com/news/anthropic-mythos-unauthorized-ac…
check_circle
“One of the unauthorized users reportedly had some level of access to Anthropic’s systems due to working as a third-party contractor for the firm.”
CORROBORATED
Multiple web search results indicate that one of the unauthorized users gained access through employment as a third-party contractor for Anthropic.
travel_explore
web search NEUTRAL — Anthropic’s Mythos Accessed by Unauthorized Users. Video Player is loading. Unmute.The users gained access to Mythos through a mix of tactics, including using access one of them had as a worker at a t…
https://www.bloomberg.com/news/articles/2026-04-21/anthropic…
travel_explore
web search NEUTRAL — ‘Handful’ of people allegedly gain unauthorised access to model adept at detecting cybersecurity vulnerabilities.“We’re investigating a report claiming unauthorised access to Claude Mythos Preview thr…
https://www.theguardian.com/technology/2026/apr/22/anthropic…
travel_explore
web search NEUTRAL — A handful of users managed to gain unauthorized access to Anthropic’s Claude Mythos – the model the company claims to be so dangerous that it would cause a wave of devastating cyberattacks if made ava…
https://nypost.com/2026/04/22/business/outsiders-breached-an…
check_circle
““We’re investigating a report claiming unauthorized access to Claude Mythos Preview through one of our third-party vendor environments,” an Anthropic spokesperson said in a statement.”
CORROBORATED
Multiple web search results confirm that an Anthropic spokesperson stated they were investigating a report of unauthorized access to Claude Mythos Preview via a third-party vendor environment.
menu_book
wikipedia NEUTRAL — Claude is a series of large language models developed by Anthropic and first released in 2023. Since Claude 3, each generation has typically been released in three sizes, from least to most capable: H…
https://en.wikipedia.org/wiki/Claude_(language_model)
menu_book
wikipedia NEUTRAL — FFmpeg is a free and open-source software project consisting of a suite of libraries and programs for handling video, audio, and other multimedia files and streams. At its core is the command-line ffm…
https://en.wikipedia.org/wiki/FFmpeg
menu_book
wikipedia NEUTRAL — Mark Fisher (11 July 1968 – 13 January 2017), also known under his blogging alias k-punk, was an English writer, music critic, political and cultural theorist, philosopher, Marxist and teacher based i…
https://en.wikipedia.org/wiki/Mark_Fisher
+ 3 more evidence sources
help
“The company added that it has no evidence the group’s unauthorized access had expanded beyond the third-party vendor’s environment or impacted any of its other systems.”
INSUFFICIENT EVIDENCE
No evidence was gathered for this claim, and no web search or cross-reference provided any information to confirm or deny the statement regarding the scope of the breach.
info
“One person in the Discord group – members of which were not named – told Bloomberg that they want to test new models rather than use them to cause chaos.”
SINGLE SOURCE
The claim that a member of the Discord group told Bloomberg they intended to test models rather than cause chaos is supported by the web search results, but no other independent sources corroborate this specific statement.
menu_book
wikipedia NEUTRAL — Discord is an American instant messaging and VoIP social platform that allows communication through voice calls, video calls, text messaging, and media. Communication can be private or in virtual comm…
https://en.wikipedia.org/wiki/Discord
menu_book
wikipedia NEUTRAL — Humam Sakhnini (born September 1968) is a management consultant, financial technology entrepreneur, and technology and video game industry executive, who has served as the chief executive officer of D…
https://en.wikipedia.org/wiki/Humam_Sakhnini
menu_book
wikipedia NEUTRAL — Midjourney is a generative artificial intelligence program and service created and hosted by the San Francisco–based "independent research lab" Midjourney, Inc. Midjourney generates images from natura…
https://en.wikipedia.org/wiki/Midjourney
schedule
“Anthropic said it shared Mythos with corporate partners — including Amazon, Google, Apple, Nvidia, CrowdStrike and JPMorgan Chase — so they could plug their own cybersecurity vulnerabilities.”
PENDING
schedule
“Prior to the rollout, Mythos broke out of a secure “sandbox” meant to restrict internet access – with a researcher only finding out “by receiving an unexpected email from the model while eating a sandwich in a park.””
PENDING
schedule
“Anthropic described the much-publicized incident as “demonstrating a potentially dangerous capability for circumventing our safeguards.””
PENDING
schedule
“Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell recently held a closed-door meeting in which they urged top bank officials to ensure their systems were ready for the risks purportedly posed by Mythos.”
PENDING
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.