Kingsoft has introduced Lingxi Pro, an autonomous office agent developed based on artificial intelligence. This product changes the approach to using AI by shifting it from an in-document function to the initial entry point for office task workflows.
At the presentation held at the ChinaJoy 2026 exhibition in Shanghai, the market leader in office software with over thirty years of experience demonstrated the new agent. Despite the gaming atmosphere of the booth with cosplayers, the essence of the product lies in its ability to accept natural language requests and output ready-to-edit files in native format.
Lingxi Pro is tightly integrated with the WPS cloud document ecosystem, which serves 678 million active devices monthly. Users can extract materials directly from cloud storage during a conversation, and everything the agent generates is saved back to the cloud documents folder, forming a closed data loop. The agent can also connect to messengers such as WeChat and Feishu, operate on cloud desktops, and support custom models in local installations, which is an unusual degree of openness for traditional office software. In this setup, the agent acts as the 'brain,' and the WPS ecosystem acts as the 'hands.'
The agent's main function is transforming skills into native files. During the demonstration, a skill for compiling an annual report, adapted for Lingxi Pro, was shown. Instead of standard output in HTML or PDF formats, Lingxi created an editable Word document and a formatted PPT presentation ready for submission. The generated files are placed in the 'Outputs' section, where Word and PPT versions are stored for later access. Furthermore, the agent supports scheduled tasks, such as a weekly process that gathers news from advanced models, summarizes it, and generates a PPT report, saving it to cloud storage, thereby automating regular reporting.
A key differentiator is the integration with cloud services. Users can view WPS documents in the cloud using a plus icon in the dialogue window, eliminating the need for manual file uploads. Documents created by Lingxi are automatically saved to the cloud in one step. For financial analysis, a single instruction is sufficient to analyze an annual bank report, after which the agent provides a categorized income and expense report with formulas next to each figure. The agent can also organize cloud drives, sorting files and drafts by version year.
The scalability potential is linked to the fact that 88% of WPS users have not yet deeply utilized AI. Of the 678 million active devices monthly, only 80.13 million, or about 12%, use AI features. Previously, the viral WPS AI feature—automatic formatting of documents with red headings for government needs—showed how a familiar tool could attract non-technical users into AI-driven workflows. However, the review noted limitations of Lingxi: the inability to directly reference local or cloud files during a conversation, and file organization based on sessions rather than projects. It is reported that Kingsoft plans to address these shortcomings in the next version, as well as add connectors for third-party resources such as research reports.
OpenAI announced on Monday the expansion of Daybreak, its cybersecurity service introduced earlier this year, and simultaneously launched a new artificial intelligence (AI) model dedicated to digital security tasks.
This initiative comes amid a growing proliferation of attacks orchestrated by AI agents, which include intrusions into Hugging Face, attacks on an academy website, and the creation of fake profiles for social engineering purposes.
Daybreak consolidates access to models, tools, and workflows intended for defense specialists. With this expansion, the service now operates at two levels, named Blue and Red, granting previously approved clients access to the company's most advanced models, which previously had considerable restrictions on permitted user use.
The frontier models are the most sophisticated available and have already generated controversy; the Trump administration, for example, sought cooperation with AI companies for the launch of these models, mentioning security-related concerns.
Critics argue that the increase in attacks carried out by AI agents also represents a marketing opportunity for the laboratories that develop such models. OpenAI framed the expanded Daybreak within this scenario, stating in a blog post:
"The world of cybersecurity is changing rapidly — threat agents will increasingly use AI to conduct cyberattacks at unprecedented speed and scale, including in fully autonomous ways. As these capabilities spread, defenders have an ever-shrinking window to prepare."
In parallel, other corporations are showing interest in hiring the protection offered by AI labs, precisely because these labs have direct knowledge of security risks. Anthropic, a competitor of OpenAI, had already made available its own cybersecurity-focused model called Mythos.
