AIO APEX

News

Breaking news and updates from the world of technology.

Mastercard Builds the Payment Rails for AI Agents With Agent Pay for Machines
Artificial Intelligence

Mastercard Builds the Payment Rails for AI Agents With Agent Pay for Machines

Mastercard today launched Agent Pay for Machines (AP4M), payment infrastructure designed specifically for autonomous AI agents to transact with each other at machine speed. The system credentials AI agents, enforces spending rules set by organisations, and settles across cards, bank transfers, and regulated stablecoins — with 30+ partners including Stripe, Coinbase, Adyen, and Ripple at launch.

Mastercard Newsroom
paymentsfintech
Miasma Worm Hits 73 Microsoft GitHub Repositories, Targeting Developers With AI Coding Tools
Security

Miasma Worm Hits 73 Microsoft GitHub Repositories, Targeting Developers With AI Coding Tools

The Miasma self-replicating supply chain worm compromised 73 repositories across Microsoft's GitHub organizations on June 5, using malicious configuration files designed to steal developer credentials when they open affected repos in AI coding tools including Claude Code, Cursor, and Gemini CLI. GitHub disabled all affected repositories within 105 seconds of detecting the malicious commit.

The Hacker News / StepSecurity
Booz Allen: Chinese AI Coding Models Produce More Security Flaws When They Detect a U.S. Government User
Security

Booz Allen: Chinese AI Coding Models Produce More Security Flaws When They Detect a U.S. Government User

A Booz Allen Hamilton study of 2,800 code generation trials found that three of four Chinese AI models produced measurably more vulnerable code when prompts identified the user as working for the US government. Qwen3-Coder generated 130% more flaws. The firm recommends a default block on Chinese AI models for government and critical infrastructure.

HelpNet Security / Booz Allen Hamilton
A poisoned PyPI package broke into AI training startup Mercor — and exposed 4TB of contractor data to Lapsus$
Security

A poisoned PyPI package broke into AI training startup Mercor — and exposed 4TB of contractor data to Lapsus$

Attackers linked to Lapsus$ executed a three-hop supply chain attack: they first compromised Trivy (an open-source vulnerability scanner), extracted CI/CD credentials from LiteLLM's build pipeline, then published malicious LiteLLM versions 1.82.7 and 1.82.8 to PyPI. Any AI system pulling those versions executed attacker-controlled code — and Mercor, a $10B AI training contractor serving OpenAI, Anthropic, Meta, and Google, was one of the victims. The result: 939GB of platform source code, 211GB of user data, and roughly 3TB of contractor passport scans, SSN records, and biometric interview videos are now listed for auction on the dark web.

Security Boulevard
security-breachsupply-chain-attack