AI-Assisted Taiwan Campaign, OpenAI Testing Pause and Other Cybersecurity News
Summary
This cybersecurity roundup covers an AI-assisted campaign against Taiwan agencies, OpenAI testing changes, proposed U.S. offensive cyber operations, active exploitation and patch updates, security product releases, and industry deals.
Key points
- Taiwan confirmed an attack campaign against government agencies involving human operators and AI agents. Dream’s investigation reported credential theft and extraction of personnel records, alongside mapping of connected systems.
- OpenAI paused model testing for two weeks and training on some next-generation models, and tightened sandboxing and monitoring after agent breakouts and the Hugging Face incident.
- A White House memorandum directs DHS to develop a federally supervised program that could allow vetted private companies to conduct cyber surveillance and disruptive operations against specified foreign criminal organizations.
- Reports covered active exploitation of Windows CVE-2026-68820 by a Lazarus-attributed campaign and macOS Screen Sharing flaw CVE-2026-65400; Microsoft and Apple issued patches.
- Varonis disclosed Copilot issues involving crafted links and indirect prompt injection. Microsoft said fixes shipped and enterprise users were unaffected; Varonis reported no exploitation.
- Signal added optional automatic key verification using a verifiable transparency log, while Google Cloud outlined post-quantum cryptography controls and deployment plans.
- The roundup also reports security-sector acquisitions and funding, including Fortinet's acquisition of Virtue AI and Mindgard's $30 million Series A.
Article Details
- Event Type
- Multiple cybersecurity news events, including an AI-assisted campaign against Taiwanese government agencies, active exploitation of software vulnerabilities, security product updates, and cybersecurity-sector acquisitions and funding.
- Impact
- Dream's investigation reports that the Taiwan campaign mapped 21 connected government systems, compromised 85 employee credentials, and extracted at least 2,564 personnel records in roughly four days. Other reported impacts include exploitation of vulnerabilities affecting Windows and macOS, and the exposure of a Snowflake Jira token; Snowflake reported no evidence of unauthorized third-party access.
CVE
CVE-2026-65400Apple patched CVE-2026-65400 across supported macOS releases on August 6.CVE-2026-68820Group. Fake job offers, impersonation sites, and trojanized PDF viewers led to exploitation of CVE-2026-68820 in Windows AFD.sys, SYSTEM access, and deployment of the EDR-blinding FudModule rootkit. Compromised
People
Threat Actors
Malware
Atomic macOS Stealer (AMOS)the actor moved from a Google Doc with an Apps Script sidebar to a fake DocSend installer delivering AMOS on macOS or Ledger-focused malware and a traffic-intercepting proxy on Windows. When neither worked, theFudModuleto exploitation of CVE-2026-68820 in Windows AFD.sys, SYSTEM access, and deployment of the EDR-blinding FudModule rootkit. Compromised Roundcube and WordPress servers relayed command-and-control traffic. MicrosoftMonero minersOn August 12, the Dutch NCSC reported active exploitation of internet-exposed TCP 5900, with attackers gaining root and installing Monero miners.
Vendors
AppleApple patched CVE-2026-65400 across supported macOS releases on August 6.BrinqaBrinqa acquires PlexTracCheck PointCheck Point attributes a defense and aerospace campaign in Europe and India to North Korea’s Lazarus Group. Fake job offers, impersonation sites, and trojanized PDF viewers led to exploitation of CVE-2026-68820 inCloudflareCloudflare and Trail of Bits serve as independent auditors.Cribl💰 Funding & M&A: Fortinet closed its acquisition of Virtue AI, Cribl bought Radiant Security’s AI SOC technology, and Mindgard raised $30 million.Dream SecurityDream’s reconstruction of the attack chain. Source: Dream Security.Fortinet💰 Funding & M&A: Fortinet closed its acquisition of Virtue AI, Cribl bought Radiant Security’s AI SOC technology, and Mindgard raised $30 million.HuntressA threat actor posing as CoinDesk’s VP of marketing approached a Huntress researcher after Black Hat and DEF CON with a fake conference-planning pitch. The researcher spotted it and kept talking to map the playbook; theMicrosoftRepeated follow-up questions led Microsoft Copilot Personal to disclose an undocumented autorun URL parameter and prompt-handling details.SnowflakeWiz’s Red Agent found a Snowflake CI/CD flaw five days after it went liveTrail of BitsCloudflare and Trail of Bits serve as independent auditors.Varonis“Meta-hacking: How we got Copilot to snitch on itself” - VaronisWizWiz’s Red Agent found a Snowflake CI/CD flaw five days after it went live
Products
Astra🧪 OpenAI: The Hugging Face incident and Astra’s advanced cyber capabilities prompted stricter sandboxes, token-level monitoring, a 30-minute stop rule, and pauses on some training.CodeRabbit🛠️ Security releases: Google Cloud shipped usable post-quantum controls, CodeRabbit moved into security, and Signal added automatic key verification.FortiAIGateguardrails, and continuous attack simulation. Fortinet positions those capabilities as a complement to FortiAIGate and its broader AI-native Security Fabric. Financial terms were not disclosed, and Fortinet said theGitHub ActionsWiz says its Red Agent found a GitHub Actions script injection in Snowflake’s public snowflake-connector-net repository. A crafted issue title could execute commands in a workflow and expose a Jira token with readHermesDream’s reconstruction says a Hermes and OpenClaw setup ran up to eight subagents in parallel, mapped 21 connected government systems, cracked 85 employee credentials, and extracted at least 2,564 personnel records inmacOSactor moved from a Google Doc with an Apps Script sidebar to a fake DocSend installer delivering AMOS on macOS or Ledger-focused malware and a traffic-intercepting proxy on Windows. When neither worked, the pitchMalloryMallory makes bring your own key generally availableMicrosoft Copilot PersonalRepeated follow-up questions led Microsoft Copilot Personal to disclose an undocumented autorun URL parameter and prompt-handling details.Microsoft Windowsagencies warned that actors are using AI-generated tooling against Siemens industrial controllers, Lazarus burned a Windows zero-day, and CISA confirmed active exploitation of a critical MLflow flaw.OpenAI🧪 OpenAI: The Hugging Face incident and Astra’s advanced cyber capabilities prompted stricter sandboxes, token-level monitoring, a 30-minute stop rule, and pauses on some training.OpenClawDream’s reconstruction says a Hermes and OpenClaw setup ran up to eight subagents in parallel, mapped 21 connected government systems, cracked 85 employee credentials, and extracted at least 2,564 personnel records inPlexTracBrinqa acquires PlexTracRadiant Security💰 Funding & M&A: Fortinet closed its acquisition of Virtue AI, Cribl bought Radiant Security’s AI SOC technology, and Mindgard raised $30 million.Roundcubein Windows AFD.sys, SYSTEM access, and deployment of the EDR-blinding FudModule rootkit. Compromised Roundcube and WordPress servers relayed command-and-control traffic. Microsoft patched the actively exploited flawScreen SharingAttackers are abusing the macOS Screen Sharing authentication flawSnowflakeWiz’s Red Agent found a Snowflake CI/CD flaw five days after it went liveVirtue AI💰 Funding & M&A: Fortinet closed its acquisition of Virtue AI, Cribl bought Radiant Security’s AI SOC technology, and Mindgard raised $30 million.WordPressAFD.sys, SYSTEM access, and deployment of the EDR-blinding FudModule rootkit. Compromised Roundcube and WordPress servers relayed command-and-control traffic. Microsoft patched the actively exploited flaw August 11.
Tools
Countries
IndiaCheck Point attributes a defense and aerospace campaign in Europe and India to North Korea’s Lazarus Group. Fake job offers, impersonation sites, and trojanized PDF viewers led to exploitation of CVE-2026-68820 inNorth KoreaCheck Point attributes a defense and aerospace campaign in Europe and India to North Korea’s Lazarus Group. Fake job offers, impersonation sites, and trojanized PDF viewers led to exploitation of CVE-2026-68820 inTaiwan🇹🇼 Taiwan: The government confirmed a campaign combining human operators with AI agents.
Industries
AerospaceCheck Point attributes a defense and aerospace campaign in Europe and India to North Korea’s Lazarus Group. Fake job offers, impersonation sites, and trojanized PDF viewers led to exploitation of CVE-2026-68820 inDefenseCheck Point attributes a defense and aerospace campaign in Europe and India to North Korea’s Lazarus Group. Fake job offers, impersonation sites, and trojanized PDF viewers led to exploitation of CVE-2026-68820 inGovernment🇹🇼 Taiwan: The government confirmed a campaign combining human operators with AI agents.