Ransomware DataBreachToday.com

Metr and Redwood Uncover Codex Safeguard Bypasses, OpenAI Says Flaws Now Patched
Third-party artificial intelligence safety evaluators found several vulnerabilities in OpenAI's monitoring system for its coding agents, showing how easy it is to bypass the monitoring. Metr and Redwood Research stress-tested OpenAI's internal misalignment monitoring system from July 21 to Aug. 11.

Microsoft Ties Storm-0940 to Automated Intrusions Across Hundreds of Targets
Microsoft linked the ConnectWise supply chain compromise to China-aligned Storm-0940, which used the open-source Artex agentic AI framework to automate credential attacks, abuse trusted cloud applications and attempt access to more than 2,250 customer tenants at machine speed.

Medical Testing Lab Told SEC Incident Likely Compromised 'Certain Patient Records'
Medical testing company Quest Diagnostics has told the U.S. Securities and Exchange Commission that it's responding to a cyber incident involving its Quanum LSM system, used by healthcare providers to place lab test orders, receive and look up lab results.

Dragos Has Learned That 'Free' Doesn't Always Mean 'Usable'
Dragos, a pioneer in the operational technology security market, and its new backer, consulting giant Accenture, are teaming up with the Sans Institute training provider and Elastic to offer $10 million worth of free OT cybersecurity support to small and rural utilities.

Test Projects Focus on AI-Enabled Drug Prescriptions, Physical Therapy Treatment
Utah has given the green light for three more healthcare technology companies to test out how their AI systems can safely provide services such as prescribing and refilling certain medications and providing AI-augmented physical therapy plans for specific health conditions.

eMudhra Says It Mistakenly Used Weak Certificates to Sign 8 Intermediary Certs
An Indian certificate authority revoked intermediate certificates after a researcher found it used a private encryption key that was only one byte in length, making it trivial for hackers to recover the key and issue their own fraudulent certs. The issue has the most salience for app developers.

Company Looks to Extend Its Cyber Resilience Roots Into Managed Detection Services
Former WatchGuard CEO Prakash Panjwani is leading Acronis' push to consolidate MSP security spending through open MDR and XDR integrations, AI-driven remediation and technician automation, with plans to expand adoption across 21,000 partners and define an acquisition strategy by early 2027.

But There's No Consensus Way to Assess Levels of Skill
A growing number of cybersecurity jobs require not just a baseline familiarity with AI tools, but significant AI expertise, according to an analysis of validated job vacancies in the field. The Sans Institute analyzed a two-day job market snapshot of 6,193 active postings.

Network and Cloud Upgrades Support NASAs Lunar and Science Missions
NASA CIO Sean Gallagher says modernization must follow mission needs, from resilient networks and shared services to cloud and financial system upgrades. He explains how the agency applies engineering review, cybersecurity and human oversight before mission-critical work.

Supply-Chain Order Sweeps in Vendor Tools Running on DeepSeek Models
Federal civilian agencies must scrub DeepSeek and any vendor tools running on its models from their networks under a U.S. Department of Homeland Security supply-chain order signed months before it surfaced Monday.

Five- to 10-Minute Attack Windows Complicated Detection and Forensic Analysis
Google linked a stealthy AOMEI supply chain attack to UNC7017, which briefly hijacked Backupper and AnyViewer update configurations to deliver Crossroad and Shinyroad malware across more than 10 countries without exploiting software vulnerabilities or compromising build pipelines.

Thompson Says Banks Must Move Beyond Blanket Blocks on Autonomous AI Agents
Transmit Security named former Socure President Matt Thompson CEO to lead its expansion into AI agent trust infrastructure, helping banks and enterprises distinguish human from autonomous traffic, govern agent authorization and enforce human oversight of high-risk transactions.

Data Theft From Back Office Services Firm Affected 62.2 Million
Plaintiffs representing about 62.2 million affected individuals reached a preliminary $80 million settlement of proposed class-action litigation against Conduent stemming from a 2024 hack on the back-office support services firm. The incident was the largest health data breach reported in 2025.

Also: Anthropic Warns That AI Agents Could Spread Across IoT Devices
In this week's ISMG Editors' Panel, four editors discussed the weaponization of food in a cyberattack against French supermarkets, the prospect of self-replicating artificial intelligence agents and why crypto agility matters beyond post-quantum cryptography.

METR Researchers Gain Employee-Level Access to Assess Unreleased Models and Publish Independent Risk Findings
Anthropic, one of the labs that pushed for embedded evaluators, said in a blog post that it is giving the independent evaluator Metr "employee-like access" to its models before deployment. This level of access is close to what Metr said are best practices for its assessors to understand how highly capable models behave.

Delaware Complaint Targets Two Patents Covering Digital Identity Authentication
Prove has sued digital identity rival KYXStart, alleging that former senior employees helped develop competing identity verification, fraud intelligence and authentication products that infringe two Prove patents, while violating contractual restrictions and undermining its competitive position.

Forrester Warns of AI-Driven Outages, Security Costs and Regulatory Risks
As AI accelerates vulnerability discovery, security teams face a flood of patches, new liability questions and rising token bills. Forrester's 2027 outlook also warns that switching models can weaken safeguards, while physical disruptions may trigger outages and reporting failures.