[ad_1]
TechCrunch reports that security vulnerabilities are more likely to exist in applications developed using artificial intelligence systems that generate code. A Stanford University study found that inappropriate and insecure programming solutions are more likely to be used among software developers given access to the AI code generation system Codex. Additionally, developers using the Codex were more likely to report insecure answers as secure compared to developers in the control group. “Code-generation systems are not currently replacing human developers. Developers who use them to speed up their tasks should carefully review the outputs and context used throughout the project,” said Neil Perry, lead co-author of the study. I’m here. But study co-author Mega Srivastava said code generation systems could be useful for low-risk tasks, such as developing exploratory research code. “In-house development company [systems]perhaps a model further trained on in-house source code may be better, as it may encourage it to produce output in line with coding and security practices,” Srivastava said. increase.
[ad_2]
Source link
