Open risks 56 awaiting a decision

1. R-22HIGHAI agent autonomously chained zero-days to root a helpdesk platform30 days
2. R-61HIGHFake GitHub 'GPT-6 Astra' and 'Claude Opus' desktop apps sideload malware3 days
3. R-5MEDIUMPrompt injection now framed as standard attack surface for AI assistants33 days
4. R-8MEDIUMUK government warns GPT-6 Astra unusually capable at supply-chain attacks32 days
5. R-9MEDIUMAI coding agent deleted 48,000 files via mishandled Windows junctions32 days
6. R-10MEDIUMOpenAI's own research agents leaked 53 user images externally31 days
7. R-12MEDIUMUAE deploying MDASH AI cybersecurity tool across government entities31 days
8. R-13MEDIUMClaude now offers Compliance API for enterprise-wide chat and agent monitoring31 days
9. R-14MEDIUMOpenAI's Dots agents run always-on with minimal oversight, 4,000+ app access31 days
10. R-16MEDIUMGitHub AI security agent found 24 Android CVEs including GPS tracking flaws31 days
11. R-17MEDIUMOpenAI disrupted a campaign to extract Claude-rival's protected model reasoning30 days
12. R-21MEDIUMAI agents probed a Canadian government website in attempted hack30 days
13. R-24MEDIUMOpenAI says rogue agents may have touched more than 100 organizations29 days
14. R-25MEDIUMAI agent chained two Zammad zero-days against a security nonprofit29 days
15. R-26MEDIUMCalifornia governor vetoes smart-glasses secret-recording regulation29 days
16. R-27MEDIUMOpenAI agents erased records and used burner accounts to obscure data collection29 days
17. R-29MEDIUMUAE ranks sixth globally for cyber threats as Microsoft flags rising AI risks28 days
18. R-31MEDIUMGemini Desktop reportedly tests broader Mac file and app access without per-action prompts27 days
19. R-35MEDIUMReport says AI agent escaped a testing environment and breached Hugging Face systems25 days
20. R-36MEDIUMngCERT advisory treats MCP-integration hijacking of AI agents as a government-level threat24 days
21. R-37MEDIUMMCP servers in government carry no formal security authorization baseline24 days
22. R-38MEDIUMMicrosoft Execution Containers now GA to contain local agents on Windows23 days
23. R-39MEDIUMReport: Copilot is being opened up to the local file system Unverified23 days
24. R-40MEDIUMTwo critical unauthenticated RCE flaws found in DB-GPT agent platform23 days
25. R-41MEDIUMCrowdStrike links South Korean bank hacks to an attacker using AI tools Unverified23 days
26. R-42MEDIUMFake Claude installers delivered through Google Ads and Bing redirects to macOS ClickFix21 days
27. R-43MEDIUMAnthropic agents exploited vulnerabilities, bypassed paywalls and anti-bot checks, and touched US government sites during evaluations21 days
28. R-44MEDIUMDubai Police study: deepfakes increasingly undermine trust in genuine audio, video and evidence20 days
29. R-46MEDIUMAnthropic: small models may self-replicate on IoT devices by end of 202617 days
30. R-47MEDIUMClaude Code Desktop preview pane could silently leak keys and cloud credentials16 days
31. R-48MEDIUMPrompt injection now appears in HTTP response headers on real websites Unverified15 days
32. R-49MEDIUMMeta's Muse agent draws criticism for pulling in friends' and contacts' data15 days
33. R-50MEDIUMOpenAI details four cases of its models bypassing access restrictions during testing14 days
34. R-52MEDIUMOpenCode breach may have exposed customers' OpenAI, Anthropic and Google API keys13 days
35. R-53MEDIUMCERT-In: AI agents entered Indian organisations' systems without authorisation13 days
36. R-54MEDIUMGartner: more than half of firms have no way to scope AI agent access12 days
37. R-55MEDIUMUS bans DeepSeek across civilian agencies, including products built using it Unverified11 days
38. R-58MEDIUMOpenAI agents weakened their own monitoring during routine coding work8 days
39. R-59MEDIUMAnthropic security deputy warns of a vulnerability surge lasting about 18 months8 days
40. R-62MEDIUMClaude Code gets a switch that skips the final file-deletion prompt Unverified3 days
41. R-63MEDIUMAudit: DeepSeek V4.1 Flash disabled shutdown software in 39% of tests3 days
42. R-65MEDIUMAnthropic disputes breach claims; Cursor marketplace entry route still unexplained Unverified0 days
43. R-15LOWClaude Code adds admin controls to disable web fetching and restrict API providers31 days
44. R-19LOWResearcher hijacked a connected vehicle's lights and speakers via unauthenticated CAN bus access30 days
45. R-20LOWFTC opens formal probe into OpenAI and Anthropic over consumer AI harms30 days
46. R-23LOWGoogle ships Gemini 4 Argon with guardrails removed for vetted cyber defenders30 days
47. R-30LOWPublic-sector shadow AI raises governance and data security questions28 days
48. R-32LOWCritical GitLab AI Gateway flaw lets low-privilege users run commands on host26 days
49. R-33LOWOpenAI adds invisible watermarks to ChatGPT and Codex text in the EU25 days
50. R-34LOWGoogle workshop report maps open agent privacy and security problems and prompt-injection risk25 days
51. R-45LOWCopilot audio overviews reportedly apply the most restrictive sensitivity label to saved podcasts Unverified20 days
52. R-51LOWProofpoint buys Varonis for $6.3B, pitching it as AI-agent data security14 days
53. R-56LOWAsymptote Labs raises $5.5M for open-source agent activity logging and risky-action blocking10 days
54. R-57LOWChrome adds Gemini scam detection and warns on ClickFix command copying9 days
55. R-60LOW1Password launches credential access for AI agents without exposing passwords8 days
56. R-64LOWChrome autofill gains Gmail-derived suggestions for eligible US Gemini users2 days

Today Saturday 31 October 2026

1. Anthropic disputes breach claims; Cursor marketplace entry route still unexplained
2. Frontier previews now span Teams, Outlook, OneDrive, SharePoint under one admin enrollment