Faster Development Demands Stronger Security

Software development is undergoing a structural shift, with boundaries between previously distinct roles becoming less rigid. “The entire software development lifecycle is changing. Nothing is staying quite the same,” Hrvoje Gorajščan, CEO of Q agency, tells ICTbusiness Media – ICTbusiness.info. He sees the transformation as an opportunity, but not as a straightforward story of programmers producing more code.

Gorajščan argues that the distribution of work among engineers is changing. Experienced developers can benefit particularly because they are better equipped to assess generated results, identify weaknesses and decide how components fit together. The value of expertise may therefore increase even as routine tasks become easier to automate. “Senior engineers have a significant advantage,” Gorajščan says.

At Q agency, some activities have accelerated considerably. Gorajščan cites examples of work being completed six or seven times faster, while warning that this figure cannot be applied to every task or interpreted without considering quality. A rapid output is not necessarily a reliable or production-ready one. “You can be much faster, but quality can also decline,” he cautions.

That trade-off changes the emphasis of software engineering. Teams need to strengthen review processes and identify problems that may not have appeared in the same form before AI-assisted development. Security becomes more important, not less, as the amount of generated code and content increases. “Security is now even more important than it used to be,” Gorajščan stresses.

A particularly sensitive issue is the transfer of customer, financial, or other confidential information to external models. Gorajščan explains that Q agency has introduced an intermediary layer designed to filter information before it reaches such systems. This helps reduce the risk of exposing data that should remain protected. “We filter what can be sent to external models,” he says.

Quality assurance must also adapt to threats associated with generative systems, including malicious instructions intended to manipulate model behaviour. Traditional software testing remains relevant but is no longer sufficient on its own. Teams must assess the security of the AI-assisted process as well as the final application.

Gorajščan’s assessment is that AI can significantly improve development efficiency if organisations maintain control over what the tools produce and what information they process. Faster delivery has commercial value, but only when it is accompanied by dependable quality, responsible engineering and stronger safeguards.