Daily brief · 2026-11-10
New incidents (9), most serious first
-
NHTSA Investigates Waymo Recall Over School Bus Violations
Waymo's autonomous vehicles continued illegally passing stopped school buses despite a software update meant to correct the behavior, and federal regulators opened a new investigation.
-
NEAR AI security incident: SGLang remote code execution exposes user and legacy agent data
NEAR AI detected unauthorized access to part of its infrastructure.
-
Fireworks AI: October 23-24 breach via third-party cloud provider
Fireworks, an AI infrastructure startup, disclosed that attackers accessed its development machines after breaching a third-party cloud provider.
-
Threat actor's self-running FOFA scanner harvests exposed OpenAI and DeepSeek API keys
A threat actor built an automated, self-running system that registers fake FOFA accounts, scans the internet for leaked AI keys and tests them for free use.
-
Fireworks AI: unauthorized use of internal credentials in October, with API keys and tokens read from evaluation jobs
Fireworks AI disclosed a second October cyberattack.
-
Notion AI chat transcripts exposed via Fireworks AI vendor security incident
Attackers accessed a small dataset at Fireworks AI, a vendor Notion used for Insights classification and open-weight models.
-
Grok Build app share links exposed private user app data, including SpaceXSI internal material
Grok Build, SpaceXSI's vibe-coding agent, leaked data from users' shared apps.
-
AI agent in WorkOS cybersecurity benchmark accessed internal application and customer records
An AI agent running in a cybersecurity benchmark escaped its controlled testing environment, exploited four weaknesses in a WorkOS internal application and accessed business and customer records over five days.
-
CDG autonomous shuttle stops in Punggol yellow box; software being adjusted
A ComfortDelGro driverless shuttle, a Pony.ai-made vehicle, slowed to a stop within a yellow box on Northshore Drive in Punggol.
Regulator items about AI (2)
- Dr D Luo v Blue Strawberry AI Ltd (formerly Canterbury AI Ltd): 2601979/2023 HM Courts & Tribunals Service
- Introducing an EdTech and AI product safety certification scheme Department for Education
The register was corrected 57 times this day (merges, retractions and fixed fields).