6 stories from 5 sources
Supporting Articles
Dozens of Microsoft-hosted software packages were found carrying a credential-stealing payload that activates automatically when opened by an AI agent. The campaign involved 73 packages and used self-replication behavior to spread the stealer once execution began.
2:34 PMArs TechnicaMeta says attackers likely took over about 20,225 Instagram accounts by exploiting a bug in its AI support chatbot. The flaw reportedly let attackers obtain account access without two-factor authentication by prompting the chatbot for a reset or recovery path.
10:31 AMThe VergeA security researcher found a critical flaw in Zcash's Orchard shielded transaction system, a privacy layer introduced in 2022 that uses zero-knowledge proofs. The issue was disclosed to the Zcash team and fixed, but it is unclear whether anyone exploited it before remediation.
1:06 PMSchneier on SecuritySummary
Dozens of Microsoft-hosted software packages were found carrying a credential-stealing payload that activates automatically when opened by an AI agent. The campaign involved 73 packages and used self-replication behavior to spread the stealer once execution began.
Why it matters
If AI agents can be reliably induced to execute malicious packages, software supply-chain attacks get cheaper, faster, and harder to contain.
Summary
Meta says attackers likely took over about 20,225 Instagram accounts by exploiting a bug in its AI support chatbot. The flaw reportedly let attackers obtain account access without two-factor authentication by prompting the chatbot for a reset or recovery path.
Why it matters
AI support tools can become high-scale account takeover infrastructure if they can be socially engineered into overriding authentication safeguards.
Summary
A security researcher found a critical flaw in Zcash's Orchard shielded transaction system, a privacy layer introduced in 2022 that uses zero-knowledge proofs. The issue was disclosed to the Zcash team and fixed, but it is unclear whether anyone exploited it before remediation.
Why it matters
When a privacy pool can be silently broken, the currency's key promise fails and past users may never be able to prove they were safe.
A curated list of other prominent stories from this day.
WhatsApp says it disrupted phishing attempts linked to NSO Group and plans legal action against the spyware maker. NSO has been blacklisted by the United States over security concerns, and WhatsApp frames the activity as targeting users through malicious lures.
10:05 AMAl JazeeraAnthropic released an initial status report on Project Glasswing, its program that offers companies help using its Mythos model to find and fix vulnerabilities in their own code. The post pushes back on the growing assumption that Mythos is uniquely strong at vulnerability discovery, arguing that claim has been overstated by uncritical coverage.
7:01 AMSchneier on SecurityRussia temporarily halted parts of its CCTV surveillance network after the killing of Iran’s Supreme Leader raised fears that AI-assisted analysis of camera feeds can be used to identify and target protected individuals. The episode sharpened concern inside Russia that ubiquitous video systems can become an assassination and espionage tool, not just a domestic security asset.
12:00 AMFinancial TimesSummary
WhatsApp says it disrupted phishing attempts linked to NSO Group and plans legal action against the spyware maker. NSO has been blacklisted by the United States over security concerns, and WhatsApp frames the activity as targeting users through malicious lures.
Why it matters
Legal pressure combined with platform enforcement can constrict the commercial spyware market and reduce the reach of high-end phishing campaigns.
Summary
Anthropic released an initial status report on Project Glasswing, its program that offers companies help using its Mythos model to find and fix vulnerabilities in their own code. The post pushes back on the growing assumption that Mythos is uniquely strong at vulnerability discovery, arguing that claim has been overstated by uncritical coverage.
Why it matters
Security buyers need evidence, not hype, because inflated claims about AI bug finding can distort tooling decisions and risk models.
Summary
Russia temporarily halted parts of its CCTV surveillance network after the killing of Iran’s Supreme Leader raised fears that AI-assisted analysis of camera feeds can be used to identify and target protected individuals. The episode sharpened concern inside Russia that ubiquitous video systems can become an assassination and espionage tool, not just a domestic security asset.
Why it matters
AI turns civilian camera infrastructure into high-value targeting infrastructure, forcing states and companies to treat video systems as critical cyber assets.
Make it yours
Pick your topics, set your cadence, and receive your personalized First Pass in your inbox. It’s that simple!