Table of Contents
AI writng code isn’t a future possibility-it’s happening now, especially in embedded software development. but this rapid change brings big questions. Who’s responsible when things go wrong? A recent report from Black Duck, teh “State of Embedded Software Quality and Safety 2025,” reveals some surprising truths about AI’s impact.
Key Findings
- 89% of teams are already using AI coding assistants.
- 75% of companies have banned AI use, yet are battling “Shadow AI”-employees using AI tools without approval.
- 30-35% of AI-generated code samples still contain vulnerabilities.
AI dramatically speeds up development, but it also introduces new security risks.This is particularly critical in embedded systems, where safety and strict regulations are paramount.
Why This Matters
Embedded software controls everything from medical devices to automotive systems. A flaw in the code can have life-or-death consequences. Simply put, speed isn’t everything. We need to address the security implications of AI-assisted coding.
Practical Guidelines for AI Integration
Don’t ignore AI, but don’t blindly trust it either. Here’s how to move forward:
- Establish a Clear AI Strategy: Define exactly where and how AI tools can be used within your development process.
- Enforce Internal Policy: Prevent Shadow AI by creating and communicating clear guidelines. Make sure everyone understands the rules.
- Treat AI Code Like an Intern’s Work: Thoroughly review and verify all AI-generated code. Static analysis and rigorous testing are essential. Don’t assume it’s correct.
- Stay Informed About Regulations: Keep up-to-date with evolving global standards, such as the EU AI Act, to ensure compliance.
As one expert noted, “Rapid development is good, but ultimately testing and verification need to be strengthened.” AI is a powerful tool,but it requires careful management and a commitment to quality.
read the full Black Duck report here.