Science & Tech · AI · 9 hrs ago
AI coding tools speed development but raise questions about software safety
AI coding assistants are helping software teams build applications faster, but speed does not show whether an application is safe or reliable.
A study of 4,867 developers found that access to an AI assistant increased completed tasks by about 26% across three experiments.
Google Cloud research also found a link between AI use and faster software delivery, alongside a link to less stable releases.
South African cloud engineering company KineticSkunk is examining how to check AI-produced code, while exploring AWS agentic technologies for its engineering services.
Its CEO, Donovan Mulder, described a prospective customer in a regulated sector whose AI-built application worked but had not been assessed for security, resilience or production readiness.
Independent engineering assessments can review code and infrastructure and test for weaknesses, helping businesses understand risks before relying on an application.
The concern is that faster development must be matched by testing, review and clear responsibility, whether code is written by AI or by people.
AI coding assistants can help developers complete tasks faster, but faster development does not prove that software is secure or ready for production.
A study of 4,867 developers found that access to an AI coding assistant increased completed tasks by approximately 26% across three field experiments.
Google Cloud’s 2025 DORA research reported a positive relationship between AI adoption and software delivery throughput, but a negative relationship with delivery stability.
KineticSkunk founder and CEO Donovan Mulder says businesses need engineering checks, testing and accountability as AI-assisted development expands.
A prospective customer in a regulated environment had built an application with AI coding tools, prompting questions about its security, resilience and production readiness.
- Who
- Software teams and businesses using AI coding tools. KineticSkunk founder and CEO Donovan Mulder discusses the risks.
- What
- AI coding tools can speed development, but their output still needs engineering review and testing.
- When
- Not stated
- Where
- South Africa is where KineticSkunk is based. The prospective customer’s location is not stated.
- Why
- Faster task completion does not establish software quality, security or readiness for production.
This story does not have two clearly opposing sides.
Google Cloud’s DORA research reported a positive relationship between AI adoption and delivery throughput, but a negative relationship with delivery stability.
Stack Overflow’s Developer Survey found that 46% of respondents distrusted the accuracy of AI-tool output, while 33% trusted it.
- Developer study
- 4,867 developers
- Productivity result
- Approximately 26% more completed tasks across three field experiments
- DORA research
- 2025
- Survey trust
- 46% distrusted AI-tool accuracy; 33% trusted it
- Company
- KineticSkunk, a South African cloud platform engineering company
1 source
What people are saying
You don’t have to stop building with AI,
I think it’s one of the most exciting developments in software engineering.
An application can function perfectly on the surface, but it might carry vulnerabilities underneath.