Why 30% of Experts Secretly Distrust AI Code

 

Image by Tyli Jura from Pixabay

Updated: June 2026.

This post has been updated to include the latest findings from Google’s annual developer research. We have integrated new statistics regarding industry adoption rates and expanded our coverage on why many senior software engineers continue to distrust AI code.

A major study from Google reveals a massive shift in the tech industry. Artificial intelligence is now a core part of daily work for software developers.

The research comes from Google's DevOps Research and Assessment (DORA) division. They gathered data from 5,000 tech professionals worldwide. According to official findings from the Google Cloud DORA Report, a massive 90% of developers now use AI tools at work. This represents a 14% increase from the previous year.

Despite this rapid adoption, a serious trust issue is brewing. Many senior developers openly distrust AI code due to errors, bugs, and security risks.

 

90% Use AI Daily ➔ Yet 30% Distrust the Output➔The Hidden Security Debt

 

The Breakdown of Trust in AI Output

Tech giants continue to push for automation. However, the people actually writing the software remain highly skeptical. The survey uncovered a massive divide in how much professionals trust the code that AI platforms generate.

• 20% of developers feel very confident in AI code.
• 46% maintain only a moderate level of trust.
• 23% trust AI-generated code only a little.
• 7% do not trust it at all.

When you combine the skeptical groups, nearly 30% of tech experts actively distrust AI code. Furthermore, 30% of those surveyed stated that AI tools showed absolutely zero improvement in their overall code quality.

Why Software Experts Are Skeptical

Ryan J. Salva leads developer tool projects at Google. He noted that while Google teams weave AI into everything from drafting documentation to refining code, human oversight is mandatory. AI models frequently suffer from "hallucinations." This is when they confidently generate flawed or outdated code blocks.

If developers blindly accept these suggestions, they introduce massive security vulnerabilities. This risk forces engineers to spend extra time debugging and auditing AI code rather than building new features.

This caution is critical because AI tools are expanding everywhere. Google is competing fiercely against Microsoft Copilot, OpenAI, and Anthropic. At the same time, coding startups like Replit and Anysphere are gaining massive traction in the developer community.

The Changing Tech Job Market

The surge in AI use comes at a difficult time for tech workers. Getting an entry-level software engineering job is tougher than ever, especially as major tech corporations continue to downsize their staff.

AI tools are meant to make software development faster and easier, not to replace human creativity entirely. Today, platforms like OpenAI's ChatGPT, Google Gemini, and Meta AI are excellent for fast research and basic code skeletons. However, they cannot replace the critical thinking required to keep software secure.

Specialized Software Verification Tools

To reduce risks, companies are turning to automated systems to double-check their developers' code. For example, cloud teams use enterprise systems like the Amazon Bedrock Models to deploy secure generative AI environments.

For routine quality control, engineering teams use automated checking platforms to catch code glitches early:

• Visual Validation: Platforms like the Applitools AI Testing Suite scan the visual layers of an application for mistakes.
• Continuous Testing: Tools like Mabl Continuous Automation run background scripts to make sure new software code does not break existing features.

Yet, even with these specialized tools in place, human experts must verify the final output before it goes live.

 

 Related Coverage: Read our comprehensive guide on the top 11 no-code AI tools for app development to discover how non-technical creators are building software without writing code in 2026. 

Why Senior Engineers Distrust AI Code

Senior software leads are noticing major mistakes in AI code every single day. While over 84% of tech teams use AI assistants to type out basic code faster, actual trust in these tools is dropping fast. 

The core issue is that large language models do not actually understand how software works. Instead, they simply guess the next most likely word or symbol based on math. This mathematical guessing game creates hidden flaws that cause entire corporate apps to crash when thousands of people try to use them at the same time.

Security Bugs Force Tech Leads to Distrust AI Code 

Automated programming tools are creating massive security risks that take way too much time to find and fix. Security audits reveal that nearly 45% of AI-generated code contains serious flaws that hackers can easily exploit. 

In fact, code review platforms recently reported a massive 322% spike in security breaches tied directly to AI files. This happens because AI tools write code that looks completely beautiful and correct on your screen, but is broken on the inside. These invisible bugs act like a needle in a haystack, forcing human developers to waste their entire workweek checking every single line to prevent a major company data leak. 

 How Tech Teams Can Fix This Problem

Engineering departments are now facing a massive bottleneck because they spend too much time cleaning up AI mistakes. Tech teams are drowning in technical debt because they waste a quarter of their work hours fixing broken code. 

Because of this extra cleaning work, final code review times have spiked by 91%, which completely wipes out the speed advantages of using AI in the first place. To survive this crisis, tech companies must stop focusing only on how fast they can generate code. Instead, managers must set strict limits on AI inputs, use automated testing tools, and force human developers to sign off on every single line of code before it goes live.

Looking Ahead

In the future, AI will assist with routine corporate tasks across almost every industry. Experts predict that over 70% of the global population will eventually embrace some form of AI assistance.

For the tech sector, AI remains a powerful assistant for code documentation, software automation, and fast drafting. However, the rise of the smart programmer who knows exactly when to distrust AI code will be the ultimate line of defense against buggy, insecure software.

  Frequently Asked Questions (FAQ)

Why do software experts distrust AI code?

Many top engineers are pushing back against automated programming. While AI tools are excellent at typing out basic templates quickly, they do not actually understand logic. Instead, they just guess words based on math. This guesswork often introduces hidden flaws that cause major software apps to crash under heavy user traffic.

What are the security risks of automated programming?

Security audits show that nearly 45% of AI-generated code contains serious flaws that hackers can easily exploit. AI assistants often write code that looks clean and correct on the surface, but includes weak data protections on the inside. These invisible bugs force human developers to spend hours checking lines to prevent company data leaks.

How can tech teams fix this coding crisis?

Companies must stop prioritizing speed over quality. To fix this bottleneck, engineering managers are setting strict limits on AI usage and introducing mandatory human code reviews. Tech teams can successfully reduce these software bugs by running every single line of automated code through rigorous testing before it ever goes live to the public.

 

 

What do you think? Do you use AI for your projects, or do you still prefer coding by hand? Let us know in the comments below!

Explore More Tech Insights: Want more analytical deep-dives, developer tips, and digital safety updates? Head over to our Home Page to browse our newest articles.