OpenAI、政府向けにサイバーセキュリティ特化GPTをデモ、モデル自体のセキュリティは誰が確保する？
OpenAIは政府関係者向けにサイバーセキュリティに特化したGPTモデルのデモンストレーションを行いました。この発表は、AIモデル自体のセキュリティ保護という重要な問題を提起しています。政府がAIを活用する際には、モデル自体の防護が誰の責任となるかが焦点となります。
関連記事
Over 17,000 games were removed from Itch.io due to pressure from an Australian activist group and payment processors like Mastercard. While the campaign aimed to combat media glorifying sexual violence, critics argue it sets a dangerous precedent for financial coercion over online creative content.
A user's Apple ID was locked after purchasing a gift card from a major retailer, effectively disabling all associated Apple services and devices. Apple's Executive Relations team is now investigating the issue after media coverage brought attention to the situation.
Ars Technica retracted an article after an AI hallucinated quotes from an open source maintainer. The maintainer was harassed by an AI agent over not merging AI-generated code. The incident involved an agentic AI instance likely using OpenClaw.
Commonwealth Bank staff used ChatGPT to find contact information for a customer, resulting in the bank accidentally disclosing personal information of one customer to another. The incident occurred when staff used unauthenticated access to the consumer ChatGPT platform to obtain phone numbers.
This week's update focuses on the delay between data breaches occurring and individuals learning about them. While companies face criminal intrusions and ransom demands, there's often a significant gap before affected people are notified.