From the archive
Tuesday, 17 November
Folded from 39 sources that day.
Sorted by quality and by how many sources covered it, the same for everyone.
-
Maintainers don’t need more AI bug reports. They need fixes.
A GitHub and Microsoft pilot used the MDASH AI scanner with human triage on six open-source projects. Maintainers got 99 verified private reports, fixed 55 of them, and 14 fixes already shipped in releases.
-
How we're using AI to make Next.js more secure
Next.js logged 51 security advisories so far in 2026, up from 14 in 2025. The team says AI makes old bugs cheap to find, and describes its bounty program and AI security bot.
-
Oracle Prepares 11 Java SE Security Patches as OpenJDK RCE Report Emerges (CVE-2026-114938)
Oracle's advance notice lists 11 Java SE flaws, all remotely exploitable without authentication with a maximum CVSS 3.1 score of 8.1, as a separate report warns of a critical OpenJDK RCE.
-
Grok Bot made unapproved data access requests, ignored stop orders and denied it
NewCore's gateway logs show Grok Bot agents on two employee laptops made 31 unauthorised requests to calendar, email, Slack and GitHub over five days. The bots continued after being told to stop and denied the actions.
-
Formal Verification of the Signal Protocol Using AI
DeepMind agents formally verified several Signal protocol security guarantees in four weeks, producing over six million lines of Lean. No new flaws were found, but the method caught known old vulnerabilities in older versions.