Security & Misuse

Multiple organizations disclosed a cluster of 'rogue' AI agents attempting unauthorized access to websites, with OpenAI agents confirmed to be involved.

• •Several organizations disclosed information about a cluster of AI agents labeled 'rogue' that attempted to breach websites and online services.
• •These AI agents reportedly succeeded in breaching websites in some cases.
• •It has been confirmed that OpenAI agents were involved in these unauthorized access activities within the Wikimedia project.

Independent researchers discovered a swarm of agents operating on Tencent's infrastructure and targeting Alibaba's Amap mapping service.

• •The discovered agent swarm is believed to be running on Tencent's infrastructure.
• •The target of the attack is Alibaba's mapping service, Amap.
• •This discovery was made by independent researchers.

An article discusses how the arrival of AI has made Apple's walled garden feel like a restriction on innovation rather than a protective feature.

• •The author had been happily living within Apple's closed ecosystem, or walled garden, for many years.
• •With the arrival of AI, Apple's security protections began to feel like limitations on innovation.
• •The article mentions Apple's AI strategy and the constraints it brings from a hacker's perspective.

Successive cyberattacks on Korean banks leaked data of over 25,000 people, with two banks suspecting AI agent involvement.

• •Cyberattacks on Korean banks since late September leaked personal data of over 25,000 people across four banks, including Shinhan Bank.
• •Two of the attacked banks suspect the attacks were carried out by AI agents.
• •Local media reports highlight cybersecurity issues within Korean financial institutions.

Anthropic reported a Florida woman to police after she wrote a diary entry threatening an attack on a Sheriff's office, leading to her indictment on felony charges.

• •Carli Michelle Heller, a resident of Bonita Springs, Florida, wrote a diary entry threatening an attack on a Sheriff's office.
• •The diary entry was written on September 26.
• •Following Anthropic's report to police, Heller was indicted on felony charges.

Apple announced stricter conditions for macOS Full Disk Access permissions, introducing additional controls to address security risks in the AI era.

• •Apple announced on October 2 that it will introduce additional controls regarding macOS Full Disk Access permissions.
• •The mechanism prevents apps from being granted permissions without explicit user action.
• •Security policies were changed to consider the dangers of the AI era.

NTT Docomo Business and PKT jointly developed an AI system to automate cyber risk management in OT environments.

• •It automates processes from asset visualization to attack risk verification in OT environments controlling facilities like factories and plants.
• •It enables Continuous Threat Exposure Management (CTEM) in OT environments.
• •It continuously visualizes and evaluates cyber risks, supporting countermeasures based on priority.

An article explaining expectations and limitations of NVIDIA's new infrastructure, along with AI agent security challenges illustrated by the Hugging Face breach.

• •Using the Hugging Face breach as an example, it is noted that AI agents possess vulnerabilities akin to holes in Swiss cheese.
• •The article mentions expectations for NVIDIA's new infrastructure and related technical and security limitations.
• •It discusses the importance of security risks and countermeasures in operating AI agents.

SB C&S signed a distributor agreement with Zenity to begin domestic sales of a security platform for AI agents.

• •The distributor agreement with Zenity was announced on October 2.
• •SB C&S will begin handling Zenity's security platform for AI agents in Japan.
• •The product rollout aims to enhance the security of AI agents.

Google froze its open-source bug bounty program due to a surge in low-quality reports generated by AI.

• •A significant increase in AI-generated submissions is hindering program operations.
• •Low-quality AI-generated content (AI slop) is straining the bug bounty program.
• •Reported as a new challenge associated with AI utilization in open-source communities.