cybersecurity

OpenSSH 10.6 disables the LZ77 coder in ssh and sshd to mitigate a CCS 2026 compression side channel, enables ssh-mldsa44-ed25519 signatures, and starts deprecating scp -R.

Senators Josh Hawley and Chris Murphy announced a bipartisan proposal to hold AI agent operators and developers liable when advanced models hack into other systems or networks.

Anthropic reports Z.ai's open-weight GLM-5.3 developed end-to-end exploits in 50 of 410 ExploitBench attempts versus 56 for Claude Mythos Preview; NIST's CAISI calls it the most cyber-capable open-weight model.

DIVD says attackers chained two Zammad zero-days (CVE-2026-102489, CVE-2026-102490) and used an automated AI agent; volunteer data was exposed.

Vercel's Next.js team released 16.3.8 and 15.5.27 fixing one high, five medium and one low severity flaw; a critical and a high fix were postponed over upstream delays.

Citrix confirmed exploitation of two critical NetScaler ADC and Gateway flaws, both CVSS v4 9.5, and shipped fixes; CISA added both to its KEV catalog as of September 27, 2026.

Vercel shipped Next.js 16.3.6 to fix a critical, 9.5-rated remote code execution flaw in ImageResponse rooted in a Satori SVG-escaping bug, CVE-2026-94545.

A bug-bounty researcher showed a paying Cloudflare Containers customer could recover up to 60 KB of leftover disk data from other customers' deleted containers; Cloudflare says it found no evidence of exploitation.

CVE-2026-90898 lets unauthenticated attackers execute arbitrary commands on Bifrost AI gateway servers via a single HTTP request; fixed in transports/v2.1.0.

The Rust security response working group and crates.io team warn that rust-lang members and popular crate owners are being lured into fake video calls to compromise their devices and accounts.

A research report ties a May 2026 RubyGems package-flooding campaign to OpenAI agents that abused a documentation-build RCE and probed a since-patched CDN key-leak bug.

A TOCTOU race in JupyterLab's update-snapshots-checkout GitHub Action let a same-second commit push bypass its authorization check and reach attacker-controlled code execution, GitHub Security Lab found.