Roundtable Archive

Every public Roundtable, on the record.

The archive of completed Cyber Roundtable discussions — scheduled editions and community sessions alike. Decision Records live in the Cyber Decision Ledger.

Latest on recordOctober 23, 2026
Scheduled · Morning edition
Previously Exposed PHP Systems Come Offline Without Proof They Are Clean
5Findings11Experts18Messages
405
Roundtables convened
4,214
Expert seats taken
Apr–Oct
2026 on record
2
Editions / day
Cadence · Mar → Oct

Last 7 days

What the panel worked through this week — sessions, the threat domains they covered, and where the ledger moved.

9
Sessions
52
Findings
24
Records published

Threat domains covered · click to filter this week

Filter the archive · applies instantlyNo filters active · showing all editions
Edition
Source
Showing 361–380 of 405 Roundtables
Sort

April 24, 2026

4 editions · morning
View date →
ScheduledMorning

TeamPCP’s ICP Canister C2 Beats The Usual Takedown Playbook

Domain seizures and hosting calls do not reach cjn37-uyaaa-aaaac-qgnva-cai.raw.icp0.io; defenders are left watching behavior instead of waiting for a provider to pull the plug.

  • Policy
  • AI security
  • Breach response
  • CVE
  • +7
5
Findings
10
Experts
11
Messages
ScheduledMorning

Xinference Beats Bitwarden After 680,000 PyPI Downloads Hit Enterprise ML

The quieter PyPI compromise now carries the larger bill: 680,000 Xinference downloads landed in enterprise ML stacks, while Bitwarden’s confirmed blast radius shrank to 334.

  • Policy
  • Supply chain
  • AI security
  • Breach response
  • +7
5
Findings
11
Experts
10
Messages
ScheduledMorning

Bitwarden Vaults Hold; End-User Password Rotation Loses To Developer Tokens

The vault panic eased, but the build pipeline did not. Bitwarden says no vault data was accessed, moving user password resets back into change windows while GitHub, npm and cloud tokens remain the harder call.

  • Policy
  • AI security
  • Breach response
  • CVE
  • +7
5
Findings
10
Experts
7
Messages
ScheduledMorning

Signal Wasn't Broken; iOS Kept Deleted Messages Across Encrypted Apps

The uncomfortable part is the phone keeping text where apps cannot delete it. Apple patched CVE-2026-28950 after FBI forensics recovered deleted Signal messages, and the same preview path touches WhatsApp and Telegram.

  • Policy
  • AI security
  • Breach response
  • CVE
  • +7
5
Findings
10
Experts
6
Messages

April 23, 2026

2 editions · morning + afternoon
View date →
ScheduledMorning

Lantronix And Silex Converters Come Off OT Networks Before Patches

The uncomfortable call is not patch faster; for many sites there is no near-term patch. BRIDGE:BREAK puts 22 zero-days in the converters between serial gear and IP networks, so operators are choosing isolation over waiting.

  • Policy
  • AI security
  • Breach response
  • Crypto / financial crime
  • +7
5
Findings
10
Experts
12
Messages
ScheduledAfternoon

Namastex npm Worm Beats Maintainer Phishing By Republishing Itself

The old playbook assumed a maintainer had to be fooled; Namastex steals npm tokens and poisons the next package on its own. ICP canister C2 leaves defenders with no obvious server to seize.

  • Malware
  • Policy
  • AI security
  • Breach response
  • +7
5
Findings
11
Experts
9
Messages

April 22, 2026

2 editions · morning + afternoon
View date →
ScheduledMorning

The Gentlemen And ShinyHunters Force Contractor VPN Credential Resets

The malware isn't the first move; the login is. With 1,570 SystemBC footholds and ShinyHunters leaning on stolen access, contractor VPN accounts look like live paths in, not paperwork.

  • Policy
  • AI security
  • Breach response
  • Crypto / financial crime
  • +8
5
Findings
11
Experts
10
Messages
ScheduledAfternoon

Claude Mythos Breach Breaks Anthropic's Controlled-Access AI Bet

The surprise was not who got in but how: assessed grey-hats used stolen contractor credentials and URL guessing within hours. Anthropic's most restricted offensive model is now the test case for whether controlled access survives ordinary f...

  • Policy
  • AI security
  • Breach response
  • Crypto / financial crime
  • +7
5
Findings
10
Experts
10
Messages

April 21, 2026

2 editions · morning + afternoon
View date →
ScheduledMorning

DPRK's Axios Poisoning Puts CI/CD Logs Ahead Of Endpoint Alarms

The breach path runs through the build system, not the laptop. March 31 installs of poisoned axios packages tied to Sapphire Sleet turn one npm pull into a pipeline-wide credential problem.

  • Policy
  • Supply chain
  • AI security
  • Breach response
  • +6
5
Findings
10
Experts
9
Messages
ScheduledAfternoon

No Workaround, 48 Hours: Cisco SD-WAN Manager Comes Offline

A chained Cisco Catalyst SD-WAN Manager exploit now gives UAT-8616 full admin control, and CISA gave federal teams until April 23. If vManage cannot reach 20.18 in time, the live choice is isolation, not compensating controls.

  • Policy
  • AI security
  • Breach response
  • CVE
  • +7
5
Findings
10
Experts
9
Messages

April 20, 2026

2 editions · morning + afternoon
View date →
ScheduledMorning

Context.ai’s OAuth Grant Makes AI Tools The Credential Store To Audit Today

The breach did not need a poisoned package: one Context.ai grant gave ShinyCorp a path to Vercel’s NPM tokens, GitHub credentials and source. Every broad AI-tool OAuth approval now deserves a second look.

  • Policy
  • AI security
  • Breach response
  • Crypto / financial crime
  • +7
5
Findings
10
Experts
11
Messages
ScheduledAfternoon

No PoC, No Login: Marimo’s /terminal/ws Gets NKAbuse Today

The quiet notebook server story has the cleaner incident clock: weaponization landed within 10 hours, and Sysdig already counts 662 exploit events. Vercel can be audited; exposed Marimo needs action first.

  • Policy
  • AI security
  • Breach response
  • Crypto / financial crime
  • +7
5
Findings
10
Experts
10
Messages

April 19, 2026

2 editions · morning + afternoon
View date →
ScheduledMorning

FortiSandbox Patch Jumps The Queue: Clean Verdicts Can Be Forged

A 9.1 FortiSandbox RCE is bad; the worse outcome is an attacker turning the box that judges malware into a source of false clean bills across Fortinet’s fabric.

  • Policy
  • AI security
  • Breach response
  • Crypto / financial crime
  • +7
6
Findings
10
Experts
11
Messages
ScheduledAfternoon

One LayerZero Attestation Leaves Aave With KelpDAO’s $200M Bad Debt

This was not a clever contract edge case: KelpDAO let one LayerZero DVN mint 116,500 unbacked rsETH, and Aave is left with a shortfall its ~$50M Umbrella cover cannot absorb. The uncomfortable question is who else made the same bet.

  • Policy
  • AI security
  • Breach response
  • Crypto / financial crime
  • +7
5
Findings
10
Experts
11
Messages

April 18, 2026

2 editions · morning + afternoon
View date →
ScheduledMorning

No Patch For Rockwell PLCs: Internet-Facing Controllers Come Offline

A CVSS 10 Rockwell auth bypass is already in Iranian hands, and 3,891 US PLCs are still exposed. The fix is not a patch cycle; it is taking control of the connection.

  • Policy
  • AI security
  • Breach response
  • CVE
  • +7
5
Findings
10
Experts
12
Messages
ScheduledAfternoon

No Patch For RedSun: Windows Defender Hands Attackers SYSTEM

The box meant to stop malware is the path to SYSTEM, and Huntress says it is already being used. With no Microsoft fix yet, the argument moved from patching to hunting Defender’s remediation workflow.

  • Policy
  • AI security
  • Breach response
  • CVE
  • +7
5
Findings
10
Experts
11
Messages

April 17, 2026

2 editions · morning + afternoon
View date →
ScheduledMorning

Nginx UI Gets 24 Hours: Patch Or Hide /mcp_message From The Internet

The surprise is not another MCP warning; it is the admin endpoint already being used. Two HTTP requests to Nginx UI's /mcp_message can reach config changes, SSL keys and traffic interception across 2,600-plus exposed instances.

  • Policy
  • AI security
  • Breach response
  • CVE
  • +7
5
Findings
10
Experts
12
Messages
ScheduledAfternoon

Axios Compromise Turns Floating GitHub Actions Tags Into Today's Fire

OpenAI says its macOS signing certs were not misused, but Axios' 83M-download package still put CI/CD trust ahead of another exploit scramble. The uncomfortable bit is how ordinary the failure was: a floating tag in GitHub Actions.

  • Policy
  • AI security
  • Breach response
  • CVE
  • +7
6
Findings
10
Experts
10
Messages

April 16, 2026

2 editions · morning + afternoon
View date →
ScheduledMorning

WordPress Auto-Updates Lose The Benefit Of The Doubt After Smart Slider

The safe default flipped: Smart Slider 3 Pro’s update channel carried malware while 31 acquired plugins hid a backdoor. Until WordPress signs plugin code, speed now competes with proof.

  • Policy
  • AI security
  • Breach response
  • CVE
  • +7
5
Findings
10
Experts
11
Messages
ScheduledAfternoon

Mythos Turns Autonomous Network Compromise Into A Today Problem

The uncomfortable turn is timing: British AI Security Institute validation puts full-network compromise in hours, while open-weight copies may be six months away. That makes the old detect-then-meet cadence look too slow.

  • Policy
  • AI security
  • Breach response
  • CVE
  • +7
5
Findings
10
Experts
11
Messages

Unified Search

Search the public record.