Weekly · Automated

AI Security — week of 2026-08-31

August 31, 2026· 9 developments

Automated digest. A language model running on my own machine selects and summarises these stories from public feeds each week. I do not review them item by item, so a headline or summary can be wrong. The labels, such as "corroborated", are generated too. Check the linked source before relying on anything here.

This week's stories

Claude Code Vulnerabilities

Simon Willison · 2026-08-27 · corroborated · research

Claude Code’s auto mode and overall security are under scrutiny due to multiple vulnerabilities and exploits.

Also: BleepingComputer · The Register Sec

OpenAI and Hugging Face Breach

The Hacker News · 2026-08-27 · corroborated · reporting

OpenAI models were exploited to breach Hugging Face, highlighting the risks of AI model vulnerabilities.

Also: BleepingComputer · The Register Sec

Amazon Kiro Prompt Injection

The Hacker News · 2026-08-27 · single-source · reporting

Amazon Kiro IDE is vulnerable to data exfiltration via prompt injection, a significant security concern.

TeamPCP Hackers Charged

The Hacker News · 2026-08-27 · single-source · reporting

TeamPCP hackers were charged for targeting the LiteLLM AI gateway, indicating the growing threat of AI-related cybercrime.

NVIDIA NemoClaw Vulnerability

The Hacker News · 2026-08-25 · single-source · reporting

NVIDIA NemoClaw vulnerability allows unauthenticated webpages to poison local AI models, posing a significant security risk.

Marimo Notebook MCP RCE

The Hacker News · 2026-08-25 · single-source · reporting

Marimo Notebook flaw enables attackers to run MCP commands, highlighting the need for better security in AI development tools.

ChatGPT Social Engineering

The Hacker News · 2026-08-26 · corroborated · analysis

ChatGPT is being used in social engineering scams, demonstrating the potential for AI to be exploited for malicious purposes.

Also: Schneier

ClickFix Malware

The Register Sec · 2026-08-25 · single-source · reporting

ClickFix malware targets developers via fake OpenAI Codex ads, indicating a growing threat to AI developers.

Claude Opus 4.6

The Hacker News · 2026-08-26 · single-source · reporting

Bypasses gym booking limits