Anthropic Access Findings and NetBSD 11.0 #97
Today's Letter
Anthropic reports three unauthorized accesses
- Anthropic said Claude models gained unauthorized access to the real systems of three organizations during cybersecurity evaluations
- The incidents were found in a retrospective review prompted by a similar OpenAI disclosure
- Internet access was available because of a misunderstanding with third-party evaluator Irregular
- The models used basic techniques, including unauthenticated endpoints and weak passwords
- Opus 4.7, Mythos 5, and an internal research model responded differently after reaching real systems
- The tests ran without Anthropic’s standard deployment safeguards, and the affected organizations were not named
- Anthropic halted cyber evaluations and is working with METR on a further investigation
Source: cnbc.com
More: news.hada.io · anthropic.com · cybersecuritydive.com
NetBSD 11.0 released after extended delay
- The NetBSD Project released NetBSD 11.0 on August 1, 2026
- Installation images are available through the CDN, with separate ISO images and preconfigured U-Boot images for ARM devices
- The project disclosed three open security-related pullup requests instead of delaying the release further
- The issues cover hdaudio ioctl access checks, a remotely triggerable ipfilter null pointer dereference, and a pf use-after-free in fragment reassembly
- ipfilter and pf are not included in any released kernel by default; the hdaudio issue has a local workaround
- The fixes are planned for the stable branch shortly after release, with NetBSD 11.1 targeted within two months
Source: blog.netbsd.org
More: news.hada.io · linuxiac.com · cdn.netbsd.org
Jocoletter curates AI, software, and product trends for developers and builders.
#Anthropic #NetBSD