Anthropic tasked its Claude Fable 5.1 model with solving Sir Thomas Urquhart's unsolved 17th-century cryptogram, the Cyphral Distich, and the AI produced a solution within about a day after roughly 44 minutes of reasoning. The cipher, two lines of 32 numbers each, had resisted human cryptographers for centuries and was even featured on Klaus Schmeh's list of top unsolved historical ciphers. Fable 5.1 succeeded by noticing that the number 32 and surrounding textual clues in Urquhart's book pointed to the decoding method that human researchers had overlooked.
vals.ai
· 2026-09-13
Anthropic's September 2026 threat report describes Chinese-based actors using its Claude models across at least five programs, including two tied to military applications. One actor drafted fire-control specifications for an anti-torpedo system and benchmarked it against U.S. Navy systems, while another built roughly 16 software modules for suppressing enemy air defenses, with a default scenario targeting 12 sites in Taiwan such as Patriot batteries and air bases. Metadata reviewed by Anthropic suggested links to Chinese research institutions, including the PLA Academy of Military Sciences.
tomshardware.com
· 2026-09-13
A new benchmark called Real-SWE evaluates frontier AI coding models against tasks drawn from real, private production codebases licensed from actual companies, including billing, tax and customer-migration work. Top performer Fable 5.1 running on Claude Code resolved 38.8% of tasks, followed by GPT-6 Astra and Gemini 3.8 Flash, with several other models trailing well below that mark.
withspecific.com
· 2026-09-12
ModelRift ran six Claude-powered agents through three printable-part tasks, split evenly between CadQuery and OpenSCAD, then independently verified every resulting STL mesh with a separate parser. All six parts printed successfully, showing both toolchains can be reliably driven by an AI agent with no human oversight. The real distinction ModelRift found wasn't whether the tools worked, but how each one behaved when it failed.
modelrift.com
· 2026-09-12
Anthropic's September 2026 threat report found that an Iran-linked group used its Claude AI model to compile targeting data on U.S. naval vessels in the Middle East, combining satellite imagery, ship transponder data and personnel names scraped from public photos, while also probing known vulnerabilities in shipboard communications gear. Separately, a Houthi-linked cell in Yemen reportedly used Claude Code to help develop guidance, navigation and control software for three weapons programs, including a guided rocket, a long-range ballistic missile, and an R2000 hypersonic glide vehicle variant. Anthropic says it banned the accounts, added detection safeguards, and reported the activity to U.S. authorities.
tomshardware.com
· 2026-09-12
New reporting this week shows Anthropic's Claude AI model being exploited for activities ranging from cyberattacks to attempts at bioweapon-related research, adding to a string of AI safety controversies. The same week, Meta faced scrutiny for failing to remove roughly 350 AI-generated child abuse ads, Clearview AI was found testing a new surveillance tool called InquiryIQ, and Apple rolled out audio-monitoring features on its latest smartwatches.
wired.com
· 2026-09-12
OpenAI announced its AI agents helped solve a 90-year-old math problem carrying a $1 million prize, building on work by mathematicians Diego Córdoba and Luis Martínez-Zoroa, though the solution still requires independent verification. Mathematician Tristan Buckmaster alleges OpenAI rushed to publish after learning of his own unpublished work with Anthropic researcher Levent Alpöge, and tried to shape who received credit. The dispute follows Anthropic's recent claim that its Claude model proved thousands of small theorems while formalizing a proof of Fermat's Last Theorem.
wired.com
· 2026-09-12
Graphify C# is a free, MIT-licensed headless indexer built on Roslyn and MSBuild that extracts semantic relationships from C# codebases, including caller-callee links, interface implementations, inheritance and overrides. Unlike text search, it resolves exact symbol bindings across overloads, generics and multiple projects, producing structured data that coding agents like Codex or Claude Code can query directly instead of guessing from string matches.
github.com
· 2026-09-12
A report from The American Prospect alleges Anthropic security officials monitor AI activists near company facilities, share threat reports with police, and track behavioral 'escalation patterns.' The claims cite a company job posting for an 'enterprise intelligence specialist' tasked with tracking threats including activism. Anthropic pushed back, telling CNET the report mischaracterizes a standard physical-security function meant to protect employees, not a surveillance program targeting activists.
cnet.com
· 2026-09-11
Anthropic disclosed that between December 2025 and August 2026 it detected and disrupted misuse of its Claude AI models by multiple threat actors, including groups tied to ShinyHunters. One French-speaking member, 'frkoo', ran automated pipelines on AWS servers that downloaded 1.8 million Android APKs, decompiled them, and scanned for hardcoded credentials using TruffleHog, funneling verified findings into a categorized Telegram network. The same actor also harvested GitHub organization emails to obtain access tokens and ran a carding site impersonating French police to sell stolen payment data.
bleepingcomputer.com
· 2026-09-11
Anthropic disclosed in a new report that it identified and stopped five instances of 'biological misuse' involving its Claude AI chatbot. The flagged cases included attempts to use the model to research dangerous viruses and investigate toxic venom for potentially harmful purposes.
gizmodo.com
· 2026-09-11
Anthropic reports identifying roughly 35 attempts between November 2025 and September 2026 by suspected state-sponsored actors trying to misuse Claude for bioweapons-related research, including queries on chikungunya gain-of-function work, bird flu adaptation for mammals, and orthopoxviruses like smallpox and mpox. At least five cases showed users evading regional access blocks through anonymizing techniques, prompting Anthropic to ban the accounts and tighten its safety systems.
techspot.com
· 2026-09-11
GoKawiil