Technology

Generative AI accelerates, complicates, and secures the future of software development

Monday, 17 November 2025 5:07AM UTC

Generative AI is transforming coding practices, boosting efficiency in some areas while introducing new security and workflow challenges, prompting a shift in developer roles and collaboration with AI tools.

Generative AI is fundamentally reshaping the landscape of software development, influencing not only how engineers write and iterate on code but also prompting a broader shift in developer mindsets and career trajectories. This transformation extends beyond coding itself into areas such as data management, monitoring, and observability, demanding that developers redefine their roles and capitalise on their unique human strengths to bridge ongoing knowledge gaps and adapt effectively to new workflows.

Large language models (LLMs), the backbone of many AI coding assistants, enable developers to brainstorm ideas, compile information, and construct code snippets more efficiently. However, these models are prone to "hallucinations," where they produce inaccurate or irrelevant information presented as fact. Early in their adoption, such hallucinations resulted in developers spending disproportionate time verifying and correcting AI-generated code, negating potential time savings. Recent advancements have seen AI tools improve their reliability by building and running tests on the code they generate, self-correcting errors and thereby reducing the frequency and impact of hallucinations.

The question of whether AI accelerates or impedes development work is nuanced. The impact on productivity varies widely depending on the developer's experience and proficiency with AI prompting. A recent study by the AI research nonprofit METR found that experienced developers working on familiar codebases actually took 19% longer when assisted by AI tools like Cursor, contradicting expectations of a 24% speed increase. These delays stem from the time required to thoroughly review and amend AI-generated code, which may not be fully prepared for specialized or complex tasks without significant human oversight. Nevertheless, even those who experienced slowdowns reported that using AI made development more enjoyable and less mentally taxing, highlighting AI’s role in alleviating cognitive load rather than simply boosting speed.

Conversely, broader industry research paints a more optimistic picture regarding AI's impact on developer productivity. Atlassian’s recent study indicated that 68% of developers save over 10 hours weekly thanks to generative AI, a marked improvement from the previous year. These saved hours are often reinvested into enhancing code quality and developing new features. Despite these gains, inefficiencies persist: half of surveyed developers reported losing significant time due to fragmented workflows, poor inter-team coordination, and difficulties accessing timely information. Notably, only a small fraction of developer time is dedicated to coding per se, with the majority spent on non-coding tasks, underscoring the need for AI tools to evolve beyond assisting purely with code generation and towards improving broader workflow integration and collaboration.

AI integration is particularly effective in site reliability engineering (SRE) and DevOps processes where telemetry data can be fed into model context protocol (MCP) servers. These platforms enable AI to reason over real-time service health metrics, logs, and error patterns without requiring exhaustive manual input. The result is a significant uplift in solving operational issues efficiently, allowing engineers to remain focused on strategic problem solving rather than routine data wrangling. This integration represents a pivotal step toward more autonomous AI workflows, where humans maintain oversight while AI handles increasingly complex and menial tasks.

Despite these promising developments, AI-generated code is not without its risks. Security remains a critical concern, with a Veracode study revealing that approximately 45% of AI-produced code contains vulnerabilities, even in widely adopted languages like Java, Python, and JavaScript. Common issues include poor safeguards against cross-site scripting and log injection attacks. The security risks are exacerbated by "vibe coding," where developers rely on AI output without explicit security requirements, potentially creating exploitable flaws at scale. Experts advocate for embedding security checks into AI-assisted workflows, leveraging AI-powered remediation tools, providing developers with secure coding training, and employing robust firewalls to mitigate these growing threats.

Moreover, AI coding assistants come with inherent limitations. Many struggle with grasping the broader project context, generating innovative solutions beyond learned patterns, and addressing highly specialised domain knowledge. Over-reliance on these tools may also lead to skill atrophy among developers and challenges in debugging AI-generated code. Ethical and operational concerns remain around intellectual property rights, data privacy, potential biases in training datasets, and the risks of vendor lock-in on proprietary AI platforms.

Ultimately, the future of software development will likely hinge on a symbiotic partnership between human developers and AI. While AI enhances ideation, offers structured design options, and automates repetitive tasks, human expertise will remain indispensable for nuanced decision-making, managing complex logic, and ensuring security and ethical standards. This human-assisted AI paradigm promises not just to transform coding practices but also to elevate the role of developers as strategic innovators within technology organisations.

📌 Reference Map:

^[1] (TechRadar) - Paragraphs 1, 2, 3, 5, 6, 7, 9, 10
^[2] (Reuters) - Paragraphs 4, 5, 6
^[3] (TechRadar) - Paragraph 4
^[5] (TechRadar) - Paragraph 8
^[4] (LLinformatics) - Paragraph 7
^[6] (Influence of AI) - Paragraph 8

Source: Noah Wire Services

More on this

https://www.techradar.com/pro/how-ai-is-changing-developers-mindsets-not-just-their-code - Please view link - unable to able to access data
https://www.reuters.com/business/ai-slows-down-some-experienced-software-developers-study-finds-2025-07-10/ - A study by the AI research nonprofit METR found that using AI coding assistants like Cursor actually slowed down experienced software developers when working on codebases they were already familiar with. Despite initially believing AI would speed up their work by 24%, developers took 19% longer to complete tasks with AI assistance. The researchers found that although AI suggestions were often directionally correct, developers had to spend time reviewing and correcting them, which led to the slowdown. This challenges the widespread assumption that AI universally boosts productivity among software engineers—a belief that has spurred significant investment in AI development tools. Previous studies have shown large productivity gains in other contexts, but METR’s findings suggest these gains do not extend to expert developers working on known code. Nonetheless, both study participants and authors continue using Cursor, noting that AI makes development more enjoyable and less mentally taxing, much like editing versus creating from scratch. The authors also noted the slowdown might not apply to junior developers or unfamiliar codebases. This nuance is important as AI continues to reshape the future of software development and the tech workforce.
https://www.techradar.com/pro/ai-is-helping-developers-save-time-but-the-struggle-to-find-timely-information-is-costing-businesses-millions - A new study by Atlassian reveals that generative AI is significantly enhancing developer productivity, with 68% of developers saving over 10 hours per week—up from 46% the previous year. These time savings are often reinvested into improving code quality and developing new features, benefiting companies overall. Despite this progress, challenges remain. Half of the developers report losing over 10 hours weekly due to inefficiencies like searching for information, and 90% lose up to 6 hours due to fragmented workflows and poor collaboration. Key barriers to productivity include difficulty accessing timely information, unclear leadership direction, and lack of inter-team coordination. The study highlights a disconnect between available AI tools, which primarily focus on coding (only 16% of a developer’s week), and the broader needs of developers, who spend 84% of their time on non-coding tasks. Atlassian’s CTO Rajeev Rajan emphasizes the need for a deeper examination of AI’s impact on developer experiences and stresses the importance of better collaboration between development teams and leadership to tackle ongoing productivity issues.
https://www.llinformatics.com/blog/ai-tools-in-software-development - AI coding assistants have several limitations, including context limitations, creativity and innovation challenges, specialized knowledge gaps, dependency on training data, and ethical and security considerations. These tools might not fully grasp the broader context of a project, leading to code suggestions that are technically correct but contextually inappropriate. They can generate code based on existing patterns but might struggle with highly innovative or creative solutions that deviate from common patterns. Additionally, they may not have deep expertise in niche domains or specialized fields, where human experts with specific knowledge are necessary. The effectiveness of AI coding assistants depends on the quality and diversity of the training data. If the training data lacks examples of specific use cases, the assistant’s code suggestions may be less relevant. Furthermore, AI-charged code can sometimes include insecure or unethical practices if such patterns were present in the training data.
https://www.techradar.com/pro/nearly-half-of-all-code-generated-by-ai-found-to-contain-security-flaws-even-big-llms-affected - A recent study by Veracode found that approximately 45% of AI-generated code contains security flaws, despite often appearing production-ready. The research analyzed over 100 large language models (LLMs) across 80 coding tasks and revealed no significant improvement in security performance among newer or larger models. Java was the most affected language, with over 70% of generated code exhibiting vulnerabilities, while Python, C#, and JavaScript showed flaw rates between 38% and 45%. Common issues included poor defense against cross-site scripting (86%) and log injection (88%). The study highlights a growing risk as AI-generated code becomes more prevalent, with companies like Google and Microsoft reportedly using AI for up to a third of their code. Veracode CTO Jens Wessling emphasized that "vibe coding," where developers rely on AI without specifying security needs, has reshaped software development. The report warns that AI can also help attackers exploit vulnerabilities at scale. To address these issues, Veracode recommends integrating security checks in AI workflows, using AI-powered remediation tools, training developers on secure practices, and deploying firewalls. Wessling stressed that security must be built into AI-driven development processes to avoid accumulating long-term security risks.
https://www.influenceofai.com/ai-coding-assistants-2025/ - AI coding assistants offer several advantages, including faster project completion times, reduced need for extensive code reviews, lower training requirements for new team members, and decreased debugging and maintenance overhead. However, they also face challenges and limitations, such as occasional generation of incorrect or inefficient code, difficulty with highly specialized or domain-specific requirements, limited understanding of complex business logic, and potential security vulnerabilities in generated code. Over-reliance on AI tools raises important questions, including the risk of skill atrophy among developers, potential loss of deep understanding of underlying systems, challenges in debugging AI-generated code, and vendor lock-in concerns with proprietary platforms. Ethical considerations include copyright and intellectual property concerns, potential bias in training data affecting code suggestions, job displacement fears within the development community, and privacy concerns with code analysis and storage.

Noah Fact Check Pro

The draft above was created using the information available at the time the story first emerged. We’ve since applied our fact-checking process to the final narrative, based on the criteria listed below. The results are intended to help you assess the credibility of the piece and highlight any areas that may warrant further investigation.

Freshness check

Score: 10

Notes: The narrative is recent, published on 16 November 2025, and appears to be original content without prior publication.

Quotes check

Score: 10

Notes: No direct quotes are present in the narrative, indicating original content.

Source reliability

Score: 10

Notes: The narrative originates from TechRadar, a reputable technology news outlet, enhancing its credibility.

Plausability check

Score: 10

Notes: The claims made in the narrative align with current industry trends and are supported by recent studies and reports.

Overall assessment

Verdict (FAIL, OPEN, PASS): PASS

Confidence (LOW, MEDIUM, HIGH): HIGH

Summary: The narrative is recent, original, and published by a reputable source. It presents plausible claims supported by current industry trends and studies, indicating a high level of credibility.

AI in software development
developer productivity
AI security risks