--- title: "#AxisOfEasy 454: Anthropic Splits Its Most Capable Model In Two — One For The Public, One For Cyber Defenders" type: "post" post_id: "33999" slug: "axisofeasy-454-anthropic-splits-its-most-capable-model-in-two-one-for-the-public-one-for-cyber-defenders" canonical: "https://axisofeasy.com/aoe/axisofeasy-454-anthropic-splits-its-most-capable-model-in-two-one-for-the-public-one-for-cyber-defenders/" markdown_url: "https://axisofeasy.com/aoe/axisofeasy-454-anthropic-splits-its-most-capable-model-in-two-one-for-the-public-one-for-cyber-defenders.md" json_url: "https://axisofeasy.com/aoe/axisofeasy-454-anthropic-splits-its-most-capable-model-in-two-one-for-the-public-one-for-cyber-defenders.json" txt_url: "https://axisofeasy.com/aoe/axisofeasy-454-anthropic-splits-its-most-capable-model-in-two-one-for-the-public-one-for-cyber-defenders.txt" published: "2026-06-12T22:00:28+00:00" modified: "2026-06-12T19:16:36+00:00" author: "Mark E. Jeftovic" categories: - "#AxisOfEasy" tags: - "account security" - "Account Takeover" - "AI safety controls" - "AI security" - "AI surveillance" - "AI-powered support tools" - "Anthropic" - "authentication bypass" - "authentication flaws" - "AWS credentials" - "Azure credentials" - "Check Point VPN" - "Claude API keys" - "Claude Fable 5" - "cloud credential theft" - "cloud infrastructure" - "cloud security" - "cloudflare" - "credential theft" - "cyber defenders" - "Cybersecurity" - "data extortion" - "data retention" - "denial of service" - "developer security" - "Discord breach claims" - "easyClaw" - "easydns" - "easynode" - "Gemini API keys" - "GitHub account takeover" - "GitHub Actions" - "Google Cloud" - "Instagram accounts" - "Instagram Reels" - "IP exhaustion" - "Malware" - "malware-as-a-service" - "Mandiant" - "Meta breach" - "Miasma malware" - "Microsoft Defender" - "model access controls" - "Mozilla" - "Mythos 5" - "Nightmare Eclipse" - "npm supply chain attack" - "OIDC tokens" - "open source ecosystem" - "OpenClaw" - "password reset abuse" - "Patch Tuesday" - "phishing awareness" - "PowerShell malware" - "privilege escalation" - "Project Glasswing" - "proof of concept exploit" - "ransomware" - "Red Hat compromise" - "SLSA attestations" - "social engineering" - "software supply chain security" - "SolarWinds Serv-U" - "student privacy" - "threat intelligence" - "TikTok scams" - "Two-Factor Authentication" - "Vidar malware" - "VPS hosting" - "vulnerability discovery" - "Windows Security" - "zero-day vulnerabilities" excerpt: "Anthropic Splits Its Most Capable Model in Two — One for the Public, One for Cyber Defenders Disgruntled Ex-Microsoft Researcher Drops Seventh Zero-Day, Targets Defender Meta Bug Exposes 20,000+ Instagram Accounts... this and more in AofE #454" site_name: "" publisher: "" language: "en-US" license: "" generator: "easyPress AI Discoverability" generator_version: "2.0.14" --- *![](https://axisofeasy.com/wp-content/uploads/2020/02/weekly-briefing-axis-of-easy.jpg)*--- *Weekly Axis Of Easy #454* ========================== --- ***Last Week’s Quote was:** “I do not know how to find out anything new without being offensive,” was by Charles Fort. No winner! **This Week’s Quote:** “Success is never final, but failure can be.” By ??? **THE RULES:** **No searching up the answer,** must be posted at the bottom of this blog post, in the comments section.* *The Prize:*** *First person to post the correct answer gets their next domain or hosting renewal on us.* --- *This is your easyDNS #AxisOfEasy Briefing for the week of June 8th, 2026. Our Technology Correspondent Joann L Barnes and easyCEO Mark E. Jeftovic send out a short briefing on the state of the ‘net and how it affects your business, security and privacy. To Listen/watch this podcast edition with commentary and insight from Joey and Len the Lengend click here.* ***In this issue:*** - **Anthropic Splits Its Most Capable Model in Two — One for the Public, One for Cyber Defenders** - **Disgruntled Ex-Microsoft Researcher Drops Seventh Zero-Day, Targets Defender** - **Meta Bug Exposes 20,000+ Instagram Accounts** - **Vidar Malware Spreads Via Fake Tutorial Videos on TikTok and Instagram** - **Red Hat npm Packages Backdoored in GitHub Account Compromise** ***Elsewhere Online:*** - **Mandiant Links Silent Ransom Group to High Value Data Extortion Campaign** - **Red Flags Raised Over Alleged 10M Discord Data Breach Notice in Maine** - **Urgent Hotfix Issued After Attackers Exploit Zero Day Vulnerability in Check Point VPN** - **Attackers Target SolarWinds Serv-U Servers to Cause Denial of Service** - **Investigation Reveals SDSU Wired Student Housing with Massive AI Surveillance Network** ![](https://axisofeasy.com/wp-content/uploads/2026/05/easynode-launch-paused-resumed-1024x723.png) [**easyClaw Launch Resumed**](https://get.easynode.ai/easyclaw) After a big lead-up and much fanfare, we launched easyClaw – the hosted openclaw VPS in last week’s edition of AxisOfEasy It was fun while it lasted, but at some point in the early hours of Saturday morning, we ran out of IPs and had to throw the brakes on sign-ups. Sorry if you were one of the people affected. We’ve since added more netblocks to the easyNode ecosystem, and we’re good to go. Try *[easyclaw](https://get.easynode.ai/easyclaw)* [**Anthropic Splits Its Most Capable Model in Two — One for the Public, One for Cyber Defenders**](https://thehackernews.com/2026/06/anthropic-releases-claude-fable-5-its.html) Anthropic launched Claude Fable 5 on June 9 as two products: Fable 5 for the public and Mythos 5 — identical but with cyber safeguards lifted — restricted to vetted defenders via Project Glasswing. Both cost $10/$50 per million input/output tokens, free on Pro, Max, Team, and Enterprise plans through June 22, then shifting to usage credits. Flagged requests are silently rerouted to the weaker Opus 4.8. Glasswing partners found vulnerabilities at record rates — Cloudflare logged 2,000 bugs, Mozilla ten times more than before — yet critical patches still average two weeks. All traffic carries mandatory 30-day retention. *More via [Thehackernews](https://thehackernews.com/2026/06/anthropic-releases-claude-fable-5-its.html)* [**Disgruntled Ex-Microsoft Researcher Drops Seventh Zero-Day, Targets Defender**](https://www.theregister.com/security/2026/06/10/nightmare-eclipse-publishes-new-windows-defender-zero-day/5253725) Security researcher Nightmare Eclipse disclosed RoguePlanet, a seventh zero-day targeting Microsoft Defender on fully patched Windows 10 and 11, hours after Microsoft’s record-breaking June Patch Tuesday. The flaw exploits a race condition for SYSTEM-level privilege escalation; public PoC code was validated by Tharros Labs analyst Will Dormann and ThreatLocker. Three of six prior zero-days — RedSun, UnDefend, and BlueHammer — were exploited before patches arrived; all six are now fixed. Microsoft’s initial response sparked infosec backlash after being interpreted as a legal threat. Nightmare Eclipse retracted a promised large-scale July 14 disclosure, citing exhaustion. *More via [Theregister](https://www.theregister.com/security/2026/06/10/nightmare-eclipse-publishes-new-windows-defender-zero-day/5253725)* [**Meta Bug Exposes 20,000+ Instagram Accounts**](https://www.infosecurity-magazine.com/news/ai-coding-adoption-governance-lags/) Meta disclosed a breach affecting 20,225 Instagram accounts after attackers exploited a verification bug in its AI-powered High Touch Support (HTS) tool, discovered May 31. HTS helps locked-out users reset passwords, but a code path bug sent reset links to unassociated email addresses, allowing attackers to access accounts lacking two-factor authentication (2FA). Exposed data included contact details, dates of birth, posts, direct messages, and linked services. Meta disabled HTS, invalidated all reset links, and enrolled affected accounts in mandatory security checkpoints. The company will fix the authentication flaw before relaunching HTS and is reviewing all account recovery flows across its platforms. *More via Infosecurity-magazine* [**Vidar Malware Spreads Via Fake Tutorial Videos on TikTok and Instagram**](https://www.infosecurity-magazine.com/news/fake-software-videos-tiktok-vidar/) Threat actors are distributing Vidar — a $300 malware-as-a-service tool that harvests credentials, financial data, and authentication tokens — through fake software tutorial videos on TikTok and Instagram Reels. ReversingLabs identified two campaigns gaming platform algorithms for maximum reach. The first used Windows-impersonating accounts with AI-voiced clips directing viewers to run PowerShell commands that silently downloaded Vidar from lookalike domain msget\[.\]run. The second baited comment replies on free Spotify Premium clips to deliver malicious download links via DM. Instagram rejected abuse reports. Organizations are advised to audit install privileges and expand phishing training to include social media. *More via [Infosecurity-magazine](https://www.infosecurity-magazine.com/news/fake-software-videos-tiktok-vidar/)* [**Red Hat npm Packages Backdoored in GitHub Account Compromise**](https://hackread.com/miasma-malware-red-hat-packages-github-account/) On June 1, 2026, Microsoft, Wiz Research, Snyk, and Aikido uncovered a supply chain attack on Red Hat’s @redhat-cloud-services npm namespace. Attackers accessed a legitimate Red Hat employee’s GitHub account, injecting malicious code into RedHatInsights repositories via unauthorized commits that exploited GitHub Actions OIDC tokens to publish backdoored packages with valid SLSA attestations. At least 32 packages and 96 versions — downloaded 80,000–117,000 times weekly — were compromised. The malware, Miasma, steals AWS, Azure, Google Cloud, SSH, and AI credentials including Claude and Gemini API keys, then self-propagates to additional packages. Administrators revoked malicious versions within hours. *More via [Hackread](https://hackread.com/miasma-malware-red-hat-packages-github-account/)* --- ***Elsewhere Online:*** **Mandiant Links Silent Ransom Group to High Value Data Extortion Campaign** *Read: https://www.darkreading.com/cyberattacks-data-breaches/silent-ransom-us-law-firms-extortion-attacks* **Red Flags Raised Over Alleged 10M Discord Data Breach Notice in Maine** *Read: https://hackread.com/maine-govt-portal-discord-data-breach-notice/* **Urgent Hotfix Issued After Attackers Exploit Zero Day Vulnerability in Check Point VPN** *Read: https://www.infosecurity-magazine.com/news/check-point-critical-auth-bypass/* **Attackers Target SolarWinds Serv-U Servers to Cause Denial of Service** *Read: https://www.securityweek.com/solarwinds-patches-exploited-serv-u-vulnerability/* **Investigation Reveals SDSU Wired Student Housing with Massive AI Surveillance Network** *Read: https://reclaimthenet.org/sdsu-adds-1300-ai-cameras-330-in-student-dorms* --- **Previously on #AxisOfEasy** If you missed the previous issues, they can be read online here: - - - - - - - - - - - [June 5th, 2026: Instagram’s AI Chatbot Exploited To Hijack High-Profile Accounts](https://axisofeasy.com/453) - [May 29th, 2026: Canada’s Bill C-22 Draws Global Tech Backlash Over Surveillance Demands](https://axisofeasy.com/352) - [May 22nd, 2026: Ontario Police Secretly Used Israeli Spyware, Watchdog Finds](https://axisofeasy.com/451) - [May 15th, 2026: Foxconn Hit by Nitrogen Ransomware, 8 TB of Client Data Stolen](https://axisofeasy.com/450) - [May 8th, 2026: Canada’s Parliament Is Filing Your Posts About Politicians](https://axisofeasy.com/449)