Metaは、社内のマウス操作追跡技術に関連するセキュリティ上の懸念について精査を行っている。このシステムは従業員の活動を記録し、AIの学習を支援する役割を担っていたが、同社はこの追跡プログラムを一時停止する方針である。
問題となっている技術は、従業員のマウスの動きを監視するものである。同技術はAIモデルの開発にも活用されてきた。
Metaは現在、このツールに関連する潜在的なセキュリティリスクを調査している。当局者は、精査が行われている間、追跡プログラムを停止することを決定した。
今回の一時停止は、社内におけるデータ取り扱いや従業員の監視慣行に対する懸念に同社が対処する中で行われたものである。
AIによるレポート AIによって生成された画像
Meta has paused its Model Capability Initiative after sensitive employee data was exposed internally. The program tracks workers' keystrokes and mouse movements to train AI models. Company officials said they are investigating the incident.
Meta workers have begun circulating flyers at US offices to protest a company program that tracks their mouse movements and keystrokes. The initiative aims to train AI models but has sparked concerns among employees about surveillance and job security.
AIによるレポート
Meta is deploying software on US employees' work computers to monitor keystrokes, clicks, mouse movements, and screenshots in work apps for AI training data. Internal memos reveal no opt-out option, sparking employee discomfort, as the company invests billions in AI while cutting thousands of jobs.
Meta CEO Mark Zuckerberg announced during the company's first-quarter earnings call that it is developing personal AI agents for its smart glasses and business use. The agents aim to help users achieve personal goals and assist entrepreneurs. This builds on the newly released Muse Spark model from Meta's AI lab.
AIによるレポート
More than 70 civil liberties and advocacy organizations, including the ACLU, EPIC, and Fight for the Future, have called on Meta to scrap facial recognition plans for its Ray-Ban and Oakley smart glasses. The groups warn the 'Name Tag' feature could empower stalkers, abusers, and law enforcement to silently identify people, endangering abuse victims, immigrants, and LGBTQ+ individuals.
Hackers exploited Meta's AI support chatbot to take over Instagram accounts by tricking it into changing associated email addresses. The vulnerability allowed password resets without two-factor authentication after matching locations via VPN. Meta resolved the issue with an emergency patch on May 29.
AIによるレポート
Workers paid to train advanced AI models are increasingly relying on chatbots like ChatGPT to generate the required conversations and tests. This shortcut, described as widespread by multiple sources, risks degrading the quality of future models through recursive training on synthetic data.