Weekly AI Digest: Open Source Shakeups, Sky-High Valuations, and Training Ethics in Question
Week 14, 2025
Welcome to Week 14 of the AI Digest!
This week’s stories are all about scale and scrutiny. From OpenAI’s funding round and open-source tease to questions about how these models are trained, this week hasn’t been as hot as other weeks (if we don’t cross into the economics and stock market world…), but that doesn’t mean we don’t have some interesting news to dive into.
Let’s do it 👇
💸 Gartner Predicts GenAI Spending Will Hit $644B by 2025
Gartner has released a new forecast estimating that global spending on generative AI will reach $644 billion by the end of 2025. Enterprise IT leaders are being called to rethink priorities and invest heavily in AI, or risk irrelevance.
At the same time, it’s utterly important to understand the key factors for successfully deploying these systems. That means tailoring them for success by ensuring smooth integration with existing infrastructure, effectively moving from proof of concept to production, and having the ability to monitor and continuously improve performance. But the most importantly, is how you can build these systems focusing on real business value.
🔓 OpenAI Teases Open Source Model Release
In a surprising announcement (or perhaps, not so), Sam Altman revealed OpenAI plans to open source one of its models.
While details are still scarce, this could signal that OpenAI is indeed trying to land on a better side of the story when it comes to open source. Or that pressure from open-source competitors like Meta (LLaMA) and Mistral are pushing the company to contribute to the ecosystem more than what they have done in the past.
If true, it would mark a major transparency win for the entire industry as OpenAI is, arguably, the most famous and powerful company in the space.
🚀 OpenAI’s Valuation Hits $300 Billion
On another note, the New York Times reports that OpenAI is now valued at $300 billion, as talks of a secondary share sale emerge.
The valuation reflects continued investor confidence in OpenAI’s dominance in foundational models and its rapid product expansion.
📚 Did OpenAI Train Models on Paywalled O’Reilly Books?
Finally, a new investigation claims OpenAI may have trained its models using paywalled O’Reilly Media books, potentially without proper licensing.
Although these concerns are not new, we have new evidence that lays the ground for discussion over copyright and consent in training data practices. The types of data and media that companies can use to train large language models remain largely unregulated, leaving a significant gray area in terms of legality and ethics.
As these models become more powerful and widely adopted, we’ll see these ethical concerns jumping to the forefront. It’s probably not fair that you can see a life’s work being entirely replicated in a question of seconds, without any compensation.
If it really gets confirmed that GPT has been trained with copyrighted books, it could mark the start of legal challenges and industry-wide pressure for transparency.
See You Next Week 👋
This week, we’ve continued our journey toward embedding AI more deeply into the fabric of society. But let’s not forget that this isn’t just another wave of technological progress. It’s a profound transformation that will reshape how we live, work, and interact with the world around us.
The future of AI goes far beyond tools and features. It’s about the values we embed in these systems, the economic shifts they trigger, and the transparency we demand from those building them. The real challenge isn’t just building smarter machines, it’s ensuring that they serve us in ways that are ethical and truly beneficial to society.
If you found this helpful, consider subscribing and sharing.
See you next week!
