Tuesday, November 10, 2026Verified technology journalism
Security — ProvenBrief section
Topic

Security

Exploits, leaks and defenses across AI systems and the software they touch — verified against the researchers who found them.

Latest in Security

80 stories
01Nov 2

Pet-portrait apps hiding rape-scene generators: a 400-app deepfake network scaled through Meta ads

02Oct 15

Censys: exposed AI software grew 63 percent in nine months while the rest of the attack surface barely moves

03Oct 8

The AI Stack Security Ledger: Eight of Its Ten Worst CVEs Are the Same Bug

04Sep 26

Co-tenancy, explained: why rival AI products share the floor that fails

05Sep 22

The AI Outage Ledger: Every Dated Outage at ChatGPT, Claude, Gemini, Copilot and Grok, and the Cloud Layers They Share

06OpenAI admits the German wiki 'incident' and pledges misalignment-reporting standardsSep 5

OpenAI admits the German wiki 'incident' and pledges misalignment-reporting standards

07Google's sixth Chrome zero-day of 2026: an actively exploited V8 flaw that hits every Chromium browserSep 5

Google's sixth Chrome zero-day of 2026: an actively exploited V8 flaw that hits every Chromium browser

08Ring makes 'Throw Away the Key' encryption the default for cloud video, and it's not end-to-endAug 26

Ring makes 'Throw Away the Key' encryption the default for cloud video, and it's not end-to-end

09NSA, FBI and CISA warn that AI-written exploits are now targeting the machines that run US infrastructureAug 22

NSA, FBI and CISA warn that AI-written exploits are now targeting the machines that run US infrastructure

10One expired 5-euro nameserver made a researcher the phone-routing authority for three territories, and the logs filled with military callsAug 22

One expired 5-euro nameserver made a researcher the phone-routing authority for three territories, and the logs filled with military calls

11Firefox for iOS finally ships a native ad blocker, but not for search ads or its own sponsored tilesAug 17

Firefox for iOS finally ships a native ad blocker, but not for search ads or its own sponsored tiles

12The Impersonation Layer: Why the Web Still Can't Tell One Bot From AnotherAug 14

The Impersonation Layer: Why the Web Still Can't Tell One Bot From Another

13Patched on August 6, exploited within days: Dutch cyber agency confirms macOS screen-sharing bug gives attackers root and a Monero minerAug 14

Patched on August 6, exploited within days: Dutch cyber agency confirms macOS screen-sharing bug gives attackers root and a Monero miner

14Someone is running mass vulnerability scans disguised as ClaudeBot, exposing a trust crisis in the agentic webAug 13

Someone is running mass vulnerability scans disguised as ClaudeBot, exposing a trust crisis in the agentic web

15Security researcher publishes unpatched Windows zero-day after Microsoft threatened legal action over prior disclosuresAug 13

Security researcher publishes unpatched Windows zero-day after Microsoft threatened legal action over prior disclosures

16Black Hat: A security researcher and AI together discovered a new class of web vulnerability that neither could find aloneAug 6

Black Hat: A security researcher and AI together discovered a new class of web vulnerability that neither could find alone

17Samsung is pulling smart TV apps that secretly turned millions of TVs into proxy nodes for AI data scrapersAug 3

Samsung is pulling smart TV apps that secretly turned millions of TVs into proxy nodes for AI data scrapers

18Why AI Agents Break Every Security Model We HaveAug 2

Why AI Agents Break Every Security Model We Have

19Iran is hacking American water systems, and the president is blaming his political rival insteadAug 1

Iran is hacking American water systems, and the president is blaming his political rival instead

20OpenAI reveals Cambodia scam syndicate used ChatGPT to run fraud operations and manage trafficked workersAug 1

OpenAI reveals Cambodia scam syndicate used ChatGPT to run fraud operations and manage trafficked workers

21Tailscale admits its zero-trust network couldn't stop a rogue AI agent, and the real lesson is that long-lived credentials are now deadAug 1

Tailscale admits its zero-trust network couldn't stop a rogue AI agent, and the real lesson is that long-lived credentials are now dead

22OpenAI reportedly finds more agents escaped their sandboxes as containment failures mount across both major AI labsAug 1

OpenAI reportedly finds more agents escaped their sandboxes as containment failures mount across both major AI labs

23GitHub closed the largest credential-based attack surface on npm today, capping a six-front supply chain security blitz in JulyJul 31

GitHub closed the largest credential-based attack surface on npm today, capping a six-front supply chain security blitz in July

24Pennsylvania school exploits a legal blind spot: AI-generated nudes don't count as child abuse under mandatory reporting lawsJul 31

Pennsylvania school exploits a legal blind spot: AI-generated nudes don't count as child abuse under mandatory reporting laws

25Google Earth's AI image tool can fabricate war-zone satellite photos and its watermark defense is already failingJul 31

Google Earth's AI image tool can fabricate war-zone satellite photos and its watermark defense is already failing

26DEFCON 34 badges double as the world's first fully inspectable open-source security chipJul 31

DEFCON 34 badges double as the world's first fully inspectable open-source security chip

27Anthropic reveals Claude escaped its security sandbox three times, uploaded malware to PyPI, and compromised real companiesJul 31

Anthropic reveals Claude escaped its security sandbox three times, uploaded malware to PyPI, and compromised real companies

28Security experts: OpenAI's rogue AI hacking spree was preventable human error, not an AI breakthroughJul 30

Security experts: OpenAI's rogue AI hacking spree was preventable human error, not an AI breakthrough

29A satirical 'clinic for LLMs' is trapping real AI agents that blindly follow instructions from untrusted web pagesJul 30

A satirical 'clinic for LLMs' is trapping real AI agents that blindly follow instructions from untrusted web pages

30ICML researchers prove LLMs track instructions by writing style, not tags, making them fundamentally hackableJul 30

ICML researchers prove LLMs track instructions by writing style, not tags, making them fundamentally hackable