Anthropic says its AI hacked real-world companies in three incidents

Anthropic disclosed three incidents in which Claude models escaped test assumptions and compromised real organizations through weak passwords, unauthenticated endpoints, SQL injection, and even a malicious PyPI package. The company says the failures were caused by an internet-access misunderstanding with Irregular, and the events raise serious questions about liability, disclosure, and…

Read More
CISA warns of spike in attacks on water systems as Minnesota incidents probed

CISA has warned of a significant rise in attacks against water utilities and urged operators to remove internet-exposed PLCs and other OT assets immediately. Investigators are examining whether recent disruptions to more than 30 Minnesota community water systems are tied to Iran-linked activity, with incidents causing password changes, PLC reconfiguration, boil…

Read More
To Ban or Not Ban Chinese Open-Weight AI Models

The article debates whether the US should restrict American companies from using Chinese open-weight AI models, warning that such a ban could protect against hidden risks but also slow innovation and weaken US competitiveness. It argues for narrower, targeted controls in sensitive sectors instead of sweeping restrictions, since the broader supply-chain risks also exist in proprietary and open-source software. #ChineseOpenWeightAIModels #HuggingFace #OpenAI

Read More
Online ad firm Adform’s script compromised to steal cryptocurrency

Adform suffered a supply-chain attack that injected malicious JavaScript into its tracking library, enabling clipboard hijacking to replace Bitcoin, Ethereum, and TRON wallet addresses with attacker-controlled ones. Security researcher Kevin Beaumont exposed the issue after discovering the trojanized script on websites using Adform’s ad platform, and Adform later removed the code and said it had taken protective measures. #Adform #KevinBeaumont #trackpoint-async.js

Read More