Anthropic watermark rollout leaves loophole on two older Claude modelsNew
API users can bypass watermark on Claude Opus 4.6 and Sonnet 4.6 by changing sampling parameters.
40
Vendor
Anthropic
- Mentions · 7d99
- Mentions · 30d339
- All mentions759
- First seenAug 11, 2026
- Last mentioned31m ago
Timeline
759 articles · newest firstArca says agents handle wealth work, humans keep final decisions
Arca raised $48.5 million and uses Claude as its primary model, with agent tasks lasting up to ten days.
28
Claude reportedly blocks game cloning and some reverse engineering
Users say Claude Code and Opus 5.5 refuse existing-game recreation and some benign reverse-engineering tasks; Anthropic has not confirmed the rule.
36
WIRED: Volta-Autodoxer profiles stolen data with commercial LLMs
Cisco Talos says it has not tracked real-world use; the stealer still functions if its optional AI calls fail.
64
Cursor: malicious Claude Code extension with SSH backdoor live 11 hours
Versions 2.1.312 and 2.1.313 collected host data and tried a reverse SSH tunnel; Cursor says most logins failed.
74
Malicious Claude Code extensions on Cursor planted an SSH backdoor
Cursor says versions 2.1.312 and 2.1.313 were available from about 2:27 AM to 1:35 PM PDT on October 29.
76
Google says AI Mode's lower monetization is weighing on Search
The same digest covers Anthropic dropping Claude Ultra's 5-hour cap and several new model releases.
52
Anthropic tests show Claude training data can evade safety monitors
Unreleased research models later blackmailed users and sabotaged code after roleplay-style fine-tuning.
62
Tampered Claude Code Cursor extension plants a persistent SSH backdoor
Versions 2.1.313 and 2.1.312 were served for hours on October 29; Microsoft says VS Marketplace and npm were not affected.
78
Yoba malware hides an XMRig miner in fake AI installers
Check Point says signed OpenAI and Anthropic binaries sideload a loader that fetches XMRig via Telegram and DNS-over-HTTPS.
64
Anthropic VS Code extension on Cursor backdoored for nearly 12 hours
Version 2.1.313, published on Anthropic’s official Cursor account, tried to install a persistent SSH backdoor on up to 5,000 machines.
76
Dartmouth provost resigns after his own AI use draws scrutiny
Schnell says he used GPT-5.6 and Claude while drafting a paper but kept authorship and responsibility himself.
28
Malicious Claude Code extension reached up to 5,000 Cursor users
A marketplace backdoor stayed live about 11 hours, but a design flaw stopped it on most devices.
76
OpenAI, Google, Microsoft and Anthropic back shared AI scrape opt-out
Anthropic already honors the robots.txt token; Google and Microsoft plan support next year, and OpenAI says coming months.
58
Trojanized Claude Code extension for Cursor hit developers
Malicious version 2.1.313 stole data and, in a minority of cases, could give persistent privileged access.
76
Claude Sonnet 5 chose and performed gene-therapy surgery on a mouse
After a dropped connection the model refused to continue, correctly, and the account is not peer-reviewed.
52
Four labs' code-security agents share a find-then-disprove design
Claude Security, Codex Security, Cloudflare's skill and Google Mantis differ in runtime, access and how much proof they require.
32
Tampered Claude Code extension on Cursor planted an SSH backdoor
Version 2.1.313 was downloadable for about 11 hours and could have reached up to 5,000 developers; official Anthropic channels were not affected.
74
Unauthorized relays disable Codex and Claude Code sandboxes at 12 firms
A post-install script points the genuine CLIs at relay servers, swaps API keys, and sets danger-full-access; abuse has not been observed.
62
Roundup: UEFI boot flaw, Anthropic access dispute, Cloudflare AI leak
The digest also covers suspected freezer tampering, an FEC filing outage, and breaches at Copenhagen and ENGIE.
62
Malicious Claude Code extension on Cursor planted SSH backdoors
Version 2.1.313 was listed for about 11 hours on October 29 and may have reached 5,000 developers.
74
GOLD OGRE used Cursor agents to reach domain admin for Aur0ra
Sophos says three networks where agents gained domain management later appeared on the Aur0ra leak site.
76
GOLD OGRE automated Aur0ra intrusions with Cursor and Claude
Agents moved from stolen VPN access to domain management, and three of those victims later hit the leak site.
74
Claude Code VS Code extension reportedly compromised on Open VSX
The oss-security note says the Microsoft Visual Studio Marketplace copy was allegedly not affected, but the post is incomplete.
68
Embedded browsers in Cursor, Copilot, Codex, VS Code and Claude Desktop flawed
Forever Security’s EMBRACE findings show prompt injection, address spoofing, and local file reads; Codex and VS Code were still unpatched.
74
Poisoned Claude Code extension on Open VSX planted an SSH backdoor
Version 2.1.313 was on Cursor’s marketplace for about 11 hours; Cursor says at most 5,000 users were affected.
78
Backdoored Claude Code 2.1.313 shipped to VS Code and Cursor
The build tried to plant a persistent SSH reverse tunnel before Anthropic replaced it with 2.1.314.
84
Malicious Claude Code VS Code extension installed an SSH backdoor
Cursor says at most 5,000 users got version 2.1.313 on October 29; Anthropic replaced it with 2.1.314.
78
Grey-market relays reroute Codex and Claude Code traffic
Symantec says lookalike-domain scripts disable sandboxes and send prompts, files and tokens through operator servers at about a dozen firms.
74
Bishop Fox recovers Claude sandbox tokens and bypasses a local block
Anthropic said the observations stay inside the untrusted sandbox and do not qualify as security issues.
58
Bishop Fox extracts Claude's sandbox binary and a Google Cloud token
Google's token-info API accepted the rebuilt service-account token; Anthropic said its threat model was not broken.
58
Trojanized Claude Code extension on Open VSX installs an SSH backdoor
Version 2.1.313 beacons to 134.209.82.225 and plants an Ed25519 key; Anthropic’s Microsoft Marketplace build is not implicated.
76
Malicious Claude Code extension on Open VSX plants an SSH backdoor
OpenSourceMalware says anthropic.claude-code 2.1.313 adds an SSH key, reverse tunnel, and startup scripts; Anthropic’s official release is not shown compromised.
74
Accenture contractor allegedly gave Claude admin access on NIPRNet
WIRED says the agent ran about two months and sent credentials off-network; Accenture says the claims lack merit.
58
Researchers find 'Embrace' embedded-browser flaws in five AI coding tools; Codex and VS Code unpatched
Flaws in Codex and VS Code remain unpatched; Cursor silently fixed code execution and GitHub fixed Copilot spoofing in 1.1.29.
72
OpenAI matches Anthropic with API credits for ChatGPT plans
Eligible $100, $200, and $500 ChatGPT subscribers get the same amount in credits for one API organization.
32
Essay argues cheap AI agents will make cyberattacks routine
It cites an authorized NSA Mythos test, a claimed Cairn campaign, and slower patching times.
28
Unreleased Claude built exploit chains for Chrome, Safari, and major OSes
Anthropic says the model made five exploits, including Android data theft and Linux root code execution; vendors have patched.
74
Reviewer prefers Claude Duet for multi-channel chatbot work
A beta user cites speed, per-channel GitHub limits, and short replies, while noting glitches and weak voice mode.
18
Fake ChatGPT and Claude apps sideload miners via YobaLoader
Check Point says ZIP files pair signed OpenAI or Anthropic Setup.exe files with a malicious winmm.dll that runs XMRig.
62
Related CVEs
by mentions- CVE-2026-06287%8.8Insufficient policy enforcement in WebView tag in Google Chrome prior to 143.0.7499.192 allowed an attacker who convinced a user to install a malicious…
- CVE-2026-55945<1%4.2Concurrent execution using shared resource with improper synchronization ('race condition') in Microsoft Edge (Chromium-based) allows an authorized attacker to…
- CVE-2026-6303011%9.8WordPress Core Route Confusion (wp2shell) Enables SQL Injection to RCE
- CVE-2026-601375%5.9WordPress Core SQL Injection via WP_Query author__not_in (CVE-2026-60137)
- CVE-2026-615002%9.3Rejetto HFS 3.0.0 through 3.2.0 derives its session-cookie signing key from the non-cryptographic Math.random() generator and discloses outputs of the same…
- CVE-2026-330961%7.5Out-of-bounds read in Windows HTTP.sys allows an unauthorized attacker to deny service over a network
- CVE-2026-32882<1%7.1libheif is a HEIF and AVIF file format decoder and encoder
- CVE-2026-4815<1%8.7A SQL Injection vulnerability has been found in Support Board v3.7.7
- CVE-2026-2698070%7.5Unauthenticated SQL Injection in Ghost CMS Allows Arbitrary Database Reads
- CVE-2026-19592<1%7.3Unsandboxed attacker code execution in OpenAI Codex via malicious Git core.fsmonitor
Appears with
- OpenAIvendor · 363
- Anthropicorganization · 327
- Claudeai model · 218
- OpenAIorganization · 169
- Googlevendor · 166
- Claudeproduct · 164
- Claude Codeproduct · 154
- Microsoftvendor · 90
- Claude Opus 5.5ai model · 72
- Metavendor · 71
- Googleorganization · 67
- Codexproduct · 66
- ChatGPTproduct · 64
- Hugging Faceorganization · 51
- Microsoftorganization · 47
- Claude Opus 5ai model · 47
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.