Tencent unveils Hy4 AI model with safeguards against cyber misuse — Anadolu Agency

Chinese technology company Tencent has unveiled a new artificial intelligence model, Hy4, with safeguards designed to reduce the risk of its use for cyberattacks and potentially harmful biological research. This was reported by Anadolu Agency.

Restrictions for cyber tasks

Tencent said that Hy4 was trained on cybersecurity data focused on defensive tasks. In particular, the model received materials for identifying and fixing software vulnerabilities, while data on the offensive exploitation of vulnerabilities were excluded.

According to the company, internal research teams conducted safety assessments of the model and found no significant concerns regarding its ability to facilitate cyberattacks in the tested areas.

More current news is available on the UA.News Telegram channel Telegram.

Model capabilities

Tencent also said that Hy4 has substantial knowledge of biology, but its ability to design open-ended experiments is more limited than that of some of the most advanced artificial intelligence models.

The model can analyze images, interpret complex charts, write program code, and perform tasks through application programming interfaces. Hy4 is already available in several Tencent cloud and workplace applications. The company plans to release the model weights and technical documentation in the coming days.

Hy4 became the next development in Tencent's portfolio of AI products after the release of the previous version in August. The company reported improvements in the model's performance in financial analysis, scientific research, and software development.