INSPECTING ARCHIVED INTELLIGENCE (OUTDATED VERSION).

Anthropic Releases Claude Fable 5 with Enhanced Cyber Safeguards

| 2026-06-10 07:37 HIGH MEDIUM
JSON
Executive Summary AI-generated
The incident data reveals a sophisticated cyber threat, with multiple entities involved in developing and deploying advanced techniques. The UK's AI Security Institute made progress toward a universal jailbreak within an initial testing window, while external partners found Fable 5 to comply with zero harmful single-turn requests on cyberattack planning, exploit development, or defense evasion. However, the company will not use this data for training or non-safety purposes and will log all human access before deleting it after 30 days unless a safety investigation or legal obligation requires holding it longer. The autonomous writing of remote code execution exploits against FreeBSD's NFS server from a 17-year-old bug was triaged as CVE-2026-4747, demonstrating the capabilities of Fable 5. The model is described as full root for an unauthenticated attacker from anywhere on the internet, with NVD noting that the stack overflow itself does not require authentication but frames kernel code execution as reachable by an attacker. The incident data also highlights Mozilla's discovery and fixing of vulnerabilities in Firefox 150 compared to earlier versions using Opus 4.6, showcasing the potential for rapid exploitation and privilege-escalation exploits built within under a day each at a few thousand dollars or less in compute.
Technical Mitigations AI-generated
• The split in the cyber classifiers between Fable 5 and Claude Mythos 5 exists because Mythos-class models find and exploit software vulnerabilities well enough that, in Anthropic's framing, handing that capability to the general public without controls would give attackers serious uplift. • The mechanism is a set of classifiers : separate AI systems that watch for misuse and jailbreak attempts. When a request trips one, Fable 5 does not refuse.
Technical Observables
kgssapi.ko
Intelligence Metadata
Actors / Malware / CVEs / Campaigns
CVE-2026-4747CVE-2026-4747
Target & Sectors
FIVE_EYES FIVE_EYES legallegal defensedefense
Incident Timeline
‎June 9
Anthropic released Claude Fable 5, its most powerful AI model yet.
general_metric 5 Fable
On June 9, Anthropic released Claude Fable 5 , the most capable model it has ever made, generally available.
organisation Claude Fable 5
On June 9, Anthropic released Claude Fable 5 , the most capable model it has ever made, generally available.
‎2026/06/10
Anthropic released Claude Fable 5, its most powerful AI yet, with cyber safeguards.
organisation AI Security Institute
External red teams found none on long-form agentic tasks either, with one caveat Anthropic states plainly: the UK's AI Security Institute made progress toward a universal jailbreak within a brief initial testing window.
financial $10 $ input tokens
Both models cost $10 per million input tokens and $50 per million output tokens, less than half the price of the earlier Mythos Preview, and Fable 5 is available through the Claude API now.
financial $50 $ output tokens
Both models cost $10 per million input tokens and $50 per million output tokens, less than half the price of the earlier Mythos Preview, and Fable 5 is available through the Claude API now.
organisation NFS
It autonomously wrote a remote code execution exploit against FreeBSD's NFS server from a 17-year-old bug, triaged as CVE-2026-4747 .
organisation NVD
Anthropic describes the result as full root for an unauthenticated attacker from anywhere on the internet; NVD's entry is more measured, noting the stack overflow itself does not require the client to authenticate, but frames kernel code execution as reachable by an attacker able to send packets to the NFS server while the kgssapi.ko module is loaded.
infrastructure 4.6
Mozilla found and fixed 271 in Firefox 150, more than ten times what it caught in Firefox 148 using the older Opus 4.6.
infrastructure Linux
The red team's N-day experiments sharpen the point: starting from nothing but a disclosed CVE and its patch, Mythos Preview built working Linux privilege-escalation exploits in under a day each, at a few thousand dollars or less in compute.
organisation CVE
The red team's N-day experiments sharpen the point: starting from nothing but a disclosed CVE and its patch, Mythos Preview built working Linux privilege-escalation exploits in under a day each, at a few thousand dollars or less in compute.
organisation Claude Mythos Preview
Why is the capability a threat The case for treating this model carefully was laid out in April, when Anthropic released Claude Mythos Preview to a limited group through Project Glasswing .
organisation Project Glasswing
Why is the capability a threat The case for treating this model carefully was laid out in April, when Anthropic released Claude Mythos Preview to a limited group through Project Glasswing .
organisation Mythos Preview
During testing, Mythos Preview identified and exploited zero-day vulnerabilities in every major operating system and every major web browser when a user directed it to.
organisation KASLR
Hard technical barriers like KASLR and W^X still raise the cost; the warning is narrower, aimed at defenses that lean on attacker patience or manual effort, and the model can now supply itself.
organisation Cloudflare
Cloudflare alone found 2,000 bugs, 400 of them high- or critical-severity.
organisation MFA
MFA and comprehensive logging stay the baseline, so a single missed patch does not become the only thing standing between an attacker and the network.
organisation Cyber Verification Program
Anthropic has opened a Cyber Verification Program that lets vetted security professionals use its models for legitimate offensive work without the cyber safeguards.
organisation Mythos
A new 30-day data retention requirement Anthropic is also changing how it handles data for Mythos-class models.
‎June 22
Anthropic plans to widen access to its AI, Mythos 5, through a trusted-access program for enterprise customers starting June 22.
general_metric 5 Fable
Anthropic plans to widen Mythos 5 access through a trusted-access program, and says that once compute capacity catches up, it aims to fold Fable 5 back into subscription plans without the usage-credit premium that kicks in after June 22.
Tactical Metrics
Metrics
infrastructure
‎4.6
Software Version
Mozilla found and fixed 271 in Firefox 150, more than ten times what it caught in Firefox 148 using the older Opus 4.6.
Metrics
infrastructure
‎Linux
Affected Product
The red team's N-day experiments sharpen the point: starting from nothing but a disclosed CVE and its patch, Mythos Preview built working Linux privilege-escalation exploits in under a day each, at a few thousand dollars or less in compute.
Metrics
financial
10,000,000
$ Input Tokens
Both models cost $10 per million input tokens and $50 per million output tokens, less than half the price of the earlier Mythos Preview, and Fable 5 is available through the Claude API now.
Metrics
financial
50,000,000
$ Output Tokens
Both models cost $10 per million input tokens and $50 per million output tokens, less than half the price of the earlier Mythos Preview, and Fable 5 is available through the Claude API now.
Intelligence Sources
The Hacker News 2026-06-10
The Hacker News 2026-06-10