IT News on September 26th: According to Wccftech, a post about "large language models cracking paid software" has attracted attention on Reddit today.

No_Ideal8394 claims that he used the unreviewed and unrestricted Qwen3.8-Flash-Next-Uncensored to successfully crack version 6.43 of Internet Download Manager ( IDM ), allowing him to bypass the trial restrictions of IDM and turn the software into a fully activated version, thereby saving $11.95 per computer per year (Note from IT: The current exchange rate is approximately 80.4 RMB). In this process, he only spent 5 euros (currently about 38.3 RMB) to rent cloud GPU and leave the task to the model to handle.

The poster stated that he provided approximately 250,000 pieces of context information about Token to AI, and allowed the model to directly analyze the local executable files. About 30 minutes later, AI provided a viable patch solution. He also mentioned that there was no manual guidance throughout the process; all that was required was to tell AI "I want this, go and fix it."

According to the screenshot information displayed, AI directly analyzes the local executable files and locates the section related to authorization verification. By modifying just one byte of the code, it is possible to remove the IDM registration check, allowing the software to be displayed as activated.

The poster acknowledges that in the past, it was not difficult to find the IDM key. However, in the AI era, using LLM to apply patches directly is expected to be much faster. The user said that they didn't have high expectations initially, but the results were surprising. They expressed difficulty in imagining what cutting-edge models will be capable of achieving in the future.

This incident has sparked discussions about the capabilities of AI's automatic analysis software. In the past, cracking tools and registrars typically required manual reverse engineering. However, with the development of large language models combined with code analysis capabilities, some researchers believe that AI may further lower the barriers to software analysis and automated modification. But this does not mean that all software can be easily cracked; commercial software usually employs various protection mechanisms such as code obfuscation, integrity checks, online authorization verification, and hardware binding.

Industry insiders also point out that AI can assist in understanding code, locating issues, or generating suggestions for program modifications, which is distinctly different from actually bypassing the licensing of commercial software. The latter not only involves technical difficulties but also encompasses issues such as software copyright, licensing agreements, and legal responsibilities.