Here's a figure that caused quite a stir in the tech world in early June 2026: according to Anthropic, the company behind the Claude assistant, its own artificial intelligence model now writes more than 80% of the code actually integrated into its systems. In other words, the AI ​​that helps millions of people code now writes the bulk of the software that powers it.

What makes this news significant isn't just the percentage. It's that Anthropic published it in a blog post warning that AI could soon start improving itself and urging us to retain the option to pause development. When a company valued at hundreds of billions of dollars raises the alarm itself, it's worth understanding why and what that means in concrete terms for a Quebec-based SME.

1. Where does this figure of 80% come from?

The figure comes from an Anthropic article titled "When AI Builds Itself," which has been picked up by several reputable media outlets such as Scientific American and Tom's Hardware. The company explains that the proportion of code written by Claude and integrated into its own systems has grown from a few percent before the release of the Claude Code tool in early 2025 to over 80% today.

Anthropic adds another telling statistic: its engineering teams reportedly deliver about eight times more code per quarter than a few years ago. The human role doesn't disappear, but it changes in nature. Humans shift from writing each line to deciding, guiding, reviewing, and validating what the machine proposes.

• The AI-generated code is still reviewed and approved before being integrated.
• The increase in productivity is real, but it comes with new supervisory responsibilities.
• Anthropic itself states that "we are not there yet" and that total self-improvement "is not inevitable".

2. “Recursive self-improvement”: what are we talking about?

The term sounds like it's from a science fiction movie, but the idea is quite concrete. Recursive self-improvement is the point at which an AI becomes good enough to design and build an improved version of itself, with less and less human intervention. Each generation helps create the next one, faster than ever before.

Anthropic isn't saying it's happened. The company says the trend is pointing in that direction and that it would be unwise to ignore it. That's why it's advocating for maintaining "the option to slow down or temporarily pause" the development of the most advanced models, until research on security and societal structures keeps pace.

Why would a company request a brake on its own product?

This is the most surprising part. Anthropic built its reputation on AI safety, and this post is consistent with that. Calling for a pause mechanism isn't about slowing things down due to a lack of trust in the technology; it's about recognizing that a rapidly advancing technology needs safeguards designed from the outset, not as an afterthought.

3. What this changes for an SME in Quebec

One might think this debate is far removed from the day-to-day operations of a company in Trois-Rivières or the Mauricie region. In reality, it directly impacts how you will work with AI in the coming years.

• AI is becoming a true production collaborator. If it already writes 80% of the code for its creator, it can very well write your emails, contracts, meeting summaries or first drafts of documents.
• Human oversight remains essential. The more AI can do, the more important the work of reviewing and validating becomes. No one should publish or send generated content without checking it.
• Security cannot be addressed after the fact. Anthropic's approach of planning safeguards before they are needed also applies to your company: AI usage policies, access management, and protection of sensitive data.

Adopting these tools without a framework is a bit like hiring a very fast employee without giving them any guidelines. This is precisely where structured IT support makes all the difference. Our managed IT services help SMEs integrate AI effectively and securely, without exposing their data.

4. A step forward to be put into perspective, not to be feared

We need to keep a cool head. The 80% figure refers to a very specific context: Anthropic's internal code, written with its own models and review processes. This doesn't mean that all the world's programming is done automatically, nor that developers have become obsolete. Human work has shifted towards design, guidance, and quality control.

For an SME, the right approach is neither panic nor indifference. It's guided curiosity. Companies that learn to use these tools methodically now will gain a significant advantage, provided they maintain a responsible human presence at each key stage.

• Start with low-risk tasks: drafts, summaries, information research.
• Clearly define what can and cannot be done using AI in your organization.
• Train your team to recognize and correct questionable answers.

Frequently Asked Questions

Is Claude really writing his own code all by himself?

Not entirely alone. According to Anthropic, Claude writes over 80% of the code integrated into its systems, but this code is still overseen, reviewed, and approved by human engineers before being put into production. The human leads, the AI ​​executes much of the work.

What is recursive self-improvement?

The idea is that an AI could design and build an improved version of itself, with less and less human assistance. Anthropic states that we are not there yet, but that the trend justifies implementing safeguards, including the ability to pause development.

Should my company be worried about using AI?

No, provided it's properly managed. The real risk for an SME isn't AI itself, but its unregulated use: sensitive data shared by mistake, unverified content, poorly controlled access. With a clear policy and good IT support, AI becomes a valuable asset.

Adopting AI with a solid strategy

Anthropic's news illustrates one thing: artificial intelligence is progressing rapidly, and even its creators are urging caution. For an SME, the winning strategy is to move forward with a plan, not improvisation. If you want to integrate tools like Claude or Copilot into your operations while protecting your data, our team can develop a tailored roadmap for you. Discover our IT services for SMEs or contact us directly through our contact to discuss your needs.