Artificial Intelligence: Ally or Challenge in Corporate Software Development?

The adoption of artificial intelligence (AI) in DevOps processes is accelerating across companies worldwide. Predictions indicate that by 2028, around 75% of enterprise software engineers will use AI-powered coding assistants—a significant leap from just 10% in 2023. This rapid growth reflects the tangible benefits of automation: teams that integrate intelligent tools are about 30% more likely to rate their deliveries as highly effective.

One highlight is the time savings: nearly a third of DevOps professionals estimate that AI tools can save their team more than 40 hours per month, freeing up an entire workweek for innovation.

Security: The Key Challenge

Despite its advantages, AI-generated code presents significant security risks. Automated code often fails to meet strict quality and protection standards. Studies show a substantial portion of code suggested by AI contains exploitable vulnerabilities, such as exposed secrets, weak input validation, insecure infrastructure scripts, or missing authentication checks. These issues leave applications open to attacks, especially if not carefully reviewed by humans.

Such risks are not merely theoretical: at least three disclosed security breaches in 2024 were traced back to AI-generated code. API security is a crucial concern, as automated implementations may expose vulnerabilities if not handled cautiously.

"Hallucinated" Packages and New Threats

A study released in June 2024 found that about 20% of software package dependencies suggested by popular large language models like CodeLlama and GPT-4 Turbo did not exist in official repositories. This opens the door for attackers to register these package names and distribute malicious code—a tactic known as slopsquatting. Furthermore, 58% of these hallucinated names appeared repeatedly, making attacks even easier to orchestrate.

Additional analyses show that around 30% of AI-generated Python code and 24% in JavaScript contain security flaws, ranging from weak random number generation to cross-site scripting threats.

Human Supervision Remains Essential

Despite all the advances, expert oversight is still irreplaceable. While AI automates many tasks, only professionals can contextualize changes and ensure compliance with business goals and regulatory standards. Surveys reveal that over 70% of specialists review AI-generated outputs at least half the time, and 19% scrutinize every suggestion.

It's also up to human teams to ensure automated solutions meet internal policies and industry requirements, as AI lacks innate understanding of these boundaries.

AI as a Partner, Not a Replacement

AI is transforming DevOps, amplifying productivity and efficiency. However, human expertise remains indispensable. Companies that combine AI power with professional supervision will be better positioned to leverage this technological revolution—boosting results while safeguarding quality and security.