AIToday

AI is becoming closed, echoing software's open-source fight

Fortune AI1d ago6 min read
AI is becoming closed, echoing software's open-source fight

Key takeaway

An influential technologist warns that artificial intelligence is becoming closed and concentrated in a few companies, reversing the open-source principles that drove decades of software progress. Unlike the open-source movement—which proved that transparency enabled community-driven innovation, security through scrutiny, and broad knowledge-sharing—today's frontier AI systems are locked away, making it impossible for users to understand how they work or audit their reasoning. The risk extends beyond software: if most future science comes to rely on opaque AI, scientific progress itself may become locked inside private companies.

Summaries like this, in your inbox every morning.

Sign up free →

3 Key Points

  • What happened

    A technology leader, reflecting on decades-old debates with free software pioneer Richard Stallman, argues that artificial intelligence is now following the opposite path—moving toward closed systems controlled by a few companies, rather than the open, shared approach that transformed software development.

  • Why it matters

    The body argues that when AI becomes closed, it risks locking away knowledge at a critical moment—while the science is still young and methods are unsettled. It also means doctors, engineers, judges, and ordinary people relying on these systems cannot fully understand how they reach their answers, trusting what amounts to an oracle they cannot examine. The author contends that openness accelerated the tech industry by spreading knowledge and training a generation of engineers; closed AI risks reversing that progress.

  • What to watch

    The author distinguishes between two types of code: the code that runs a model (which some companies now release) and the code that built it plus the data it learned from (which companies hold back). Even the release of runnable code is framed as a courtesy with no commitment—companies make no promise to keep doing so for their most capable systems tomorrow.

FAQ

What is the difference between open and closed AI models that the author describes?
Open models (released by some Chinese labs and American companies) provide the code to run them, but hold back the code that built them and the data they learned from. The author calls this 'magic numbers you can run but cannot explain'—useful to deploy, but you cannot see how they were created. Closed frontier models are locked completely.
Why does the author say closed AI is a problem for trust and accountability?
When a company controls what a model can or cannot do and shapes how it reaches answers, anyone relying on it cannot fully understand the results. An explanation the model provides is a story assembled after the fact, not a faithful record of the computation. Ask the same question next year and you may get a different answer with no way to know whether the world changed or the vendor did.
What does the author say about the safety argument for keeping AI closed?
The author acknowledges the argument is not entirely crazy—releasing software is not like publishing a research paper, since the software itself is the capability. However, the author argues the conclusion does not follow: science itself could be classified under the same logic, yet it stays open and is instead monitored with rules. Closed models leak and get jailbroken anyway, and their concentration creates its own danger.

Discussion

No comments yet. Be the first to share your thoughts!

Log in to join the discussion

Related Articles

Stay ahead with AI news

Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.

Get Started Free

Free · takes 30 seconds · unsubscribe anytime

1 minute a day. The AI essentials.

200+ sources · Email / LINE / Slack

Get it free →