Hugo Face was identified as having been affected by internal data sets and service vouchers following a network invasion last week. The company disclosed that the attackers were using security loopholes through uploading to the Platform ' s data sets, enforcing malicious code on the server, and subsequently increasing their access to the wider internal system.
The attack was launched from the data set entrance.
In a blog, the company stated that the invasion was not a theft from traditional employee accounts, but rather that the attackers had used the platform-hosted data sets to trigger a loophole. Additional privileges are obtained after the malicious code is running on the server.
Hugo Face stated that the loopholes used in the attack had been repaired and the documents interviewed had been cancelled and rotated. At the same time, the company reminds users that if they have kept the key on the platform, they should also replace it as soon as possible and check the accounts for unusual activity.
Client data subject to confirmation
The company indicated that it was still investigating whether the incident involved the theft of data from customers or partners and that no clear conclusions had been reached. Hugo Face has reported the incident to law enforcement and has introduced a cybersecurity forensics team to assist in the investigation, while reviewing existing security measures.
In terms of disclosure, the extent of the impact of the event has been confirmed to cover internal data sets and service vouchers, but the impact on external data remains undetermined.
The company claims that outside AI agents were involved in the attack.
Hugo Face attributed the invasion to an external AI agent, who alleged that he had conducted thousands of operations in a large number of short-lived sandboxes and had built a chain of command and control through public services. However, TechCrunch states that the company was asked for evidence that it did not provide immediate support in time.
The company also indicated that the ACSE discovered the attack and used the AI model to analyse the server logs for recording the attack. Hugo Face stated that it had initially attempted to use a front-line model from a commercial supplier, but that the analysis process was affected by the security restrictions imposed by the other party, and that the log was then processed using locally deployed large-language models to avoid uploading sensitive attack records to external servers.
Additional information:The article mentions that security researchers have repeatedly criticized some of the front-line models for being too restrictive in the context of cyber-security and for being intercepted for defence and evidentiary purposes.