U.S. artificial intelligence (AI) companies OpenAI and Anthropic are facing regulatory investigations over security incidents involving AI agents and controversy over harm to users. After the Federal Trade Commission (FTC) demanded that the two companies submit relevant documents, governments in eight U.S. states also sued OpenAI, raising concerns about ChatGPT’s safety and consumer protection.According to the U.S. online news outlet Axios on October 14 local time, the FTC sent civil investigative demands (CIDs) to OpenAI and Anthropic that day. Each document was more than 50 pages long and reportedly demanded information on recent security incidents, companies’ safeguards, privacy policies, and marketing materials for related products.
A CID is an administrative procedure used by authorities to obtain materials needed for an investigation. The FTC is examining possible violations of consumer protection laws and plans to summon senior executives from both companies to testify about the effects of AI products on U.S. consumers. Since September, the FTC has been investigating safety issues at AI companies, including OpenAI and Anthropic.The investigation comes amid a series of incidents in which AI agents have accessed external systems beyond their intended scope or taken unintended actions.In July, OpenAI disclosed that, during an internal cybersecurity assessment, one of its AI models had bypassed controls in an isolated environment and infiltrated the systems of Hugging Face, an external AI platform. According to OpenAI, the model accessed Hugging Face’s servers using stolen credentials and exploiting a previously unknown vulnerability. The company said it subsequently investigated the incident and strengthened its security measures.OpenAI later disclosed another case in which an AI agent operating in a research and evaluation environment posted 53 images provided by ChatGPT users to an external image-hosting site. The images were posted through links that were not publicly indexed, but could still be accessed via the links. OpenAI said it had removed most of the images and was taking steps to remove the remaining content.It also emerged that an OpenAI agent had inadvertently accessed a web portal containing Australian government health statistics. The Australian government said no personal information had been viewed on the portal. OpenAI said its model had taken unintended actions.Anthropic has also recorded cases of its AI model making unusual use of external websites. During an internal evaluation on July 18, Anthropic’s Claude Haiku 4.5 submitted a false tip to the Philadelphia police website for tips on long-unsolved homicide cases. The model filled out a form with information written as if it knew about a real case, but the tip was classified as spam and was not forwarded to police investigators. Anthropic disclosed the incident and said it would improve safeguards in its internal evaluation process.Legal pressure is also mounting. According to Reuters, governments in eight U.S. states, including Iowa, Kansas, and Arizona, each filed lawsuits against OpenAI on October 14. They alleged that ChatGPT encourages users, including teenagers, to harm themselves or others, and that the company misled consumers in promoting the safety of its products.The states that filed suit are seeking penalties and damages for alleged violations of consumer protection laws. Some are also seeking measures to limit the development of certain AI models until their safety can be demonstrated. These are allegations by the plaintiffs, however, and no court has determined that the law was violated.
“There’s a Drink Like This?”… Chinese Beverage With “Korean” in Its Name Rakes in 85 Million Won in Monthly Sales
As AI agents increasingly access websites and perform tasks on users’ behalf, incidents are mounting in which they affect real users and external organizations, going beyond companies’ internal testing environments. With investigations by U.S. authorities and lawsuits by state governments underway, debate over AI companies’ safeguards and responsibility for consumer protection is also expected to intensify.
This content was produced with the assistance of AI translation services.
© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.
Click Here For The Original Source.
——————————————————–
