Copyright suits and state AI picks: is smart enough?── Reading training material, routing and drug-AI money by where they come from
Download the video (MP4, 10 MB)
Is an AI worth choosing only because it is smart?The answer is no. This video shows what else decides the choice. People who own the rights to songs have taken an AI developer to court. On the same day, a national agency was reported to have narrowed the AI it allows. When drug makers choose AI, the same question comes back to them. What did the system learn from?
Which path did its answer take?Whose money keeps it running?Being able to explain that is starting to decide which AI gets picked. Let us follow the day's events in that order.
01Music copyright suits against AI
Source Billboard ／ Law Commentary
We begin with songs. What an AI learned from is now being tested in court.
| Report | Defendant | Details |
|---|---|---|
| Billboard | Anthropic | Sued for training AI on Elvis Presley songs without permission |
| Law Commentary | Suno and Anthropic | Claim of about $340 million over songs by Madonna, Johnny Cash, others |
Both based on headlines; article details not verified by the site
The ones suing hold the rights to the music. From the headlines, their point is that songs were fed to a model without consent. Until now, what a developer learned from stayed an internal matter. Now it must be explained to outsiders. I think the plaintiffs want more than money. They want to know how their work was used.
The dispute is not about the answers a model gives. It is about an earlier step. Was consent obtained before learning began? The sum looks heavy for one company. That pressure will push other developers to rethink how they gather material, and to record where it came from.
Whether that material was gathered properly will become one test when we judge an AI.
Pharma is not exempt. If you train an in-house model on internal documents or papers, the same question returns to you. Confirm you hold the rights before you start. Are your own files being read by outside systems without your knowledge?
If the material is in question, the way users choose AI changes too. Next, a nation acting as a user.
02Governments start choosing AI
Source Yahoo ／ Unite.AI ／ The Daily Tar Heel
On the same day the makers were held to account, users moved too. Government offices began deciding for themselves which AI they may use.
| Who | Action | Scope |
|---|---|---|
| Swedish government agency | Blocked ChatGPT and Claude | Microsoft Copilot still allowed |
| UK | £100 million AI biosecurity institute | With OpenAI and Anthropic |
| 25 states incl. North Carolina | Asked Congress to regulate AI | States acted in step |
One country's agency stopped popular chat assistants. It kept only the assistant built into its office software. I doubt the survivor won on smarts. It sits inside systems the agency already runs, so it is easier to control. The reports do not give a reason. My guess is that the deciding point was whether the agency could watch where its information goes. Elsewhere, local governments are pressing for national rules.
Inviting developers into the research suggests a wish to curb misuse while models are still being built. States are no longer only picking tools. They are stepping in between makers and users.
In a drug company, if staff each use any AI they like, patient data or unpublished trial results could leak. So the company should name the AI it allows, and record why. Keep someone in charge of reviewing the usage rules after the choice, so decisions stay steady.
But what if another AI was answering behind the one your company chose? Next, where requested work is really processed.
03AI that routes the work
Source Forkast News ／ https://tech-insider.org/
The AI you talk to on screen may no longer be the one producing the answer.
Reports say work sent to one company's assistant is quietly passed to another company's model. To the user, nothing seems to change. Yet a different firm writes the answer. The report argues that the advantage has moved from raw intelligence to deciding where work goes. If so, ranking models by smartness will matter less and less.
| Model | Price |
|---|---|
| DeepSeek | $0.15 |
| Claude | $4 |
Comparison by tech-insider.org. The site's article says it stresses DeepSeek's large price advantage
Price will weigh heavily in that choice. The two costs compared differ by more than an order of magnitude. With a gap like that, a front desk will want to send work to the cheaper side. The problem is that users can neither know nor choose who answered.
For pharma, the danger is losing track of which company's system handled your documents. With patient information, you must be able to say which firm, in which country, processed it. Before signing, ask the vendor to disclose where work goes, and to warn you before changing it. Choosing on price alone pushes that check aside.
While routing and price unsettle what an AI is worth, big money is flowing to firms that design medicines with AI. Next, what those sums really mean.
04Money into drug-discovery AI
Source Wowtale ／ Kingy AI ／ BioSpectrum Asia
A big figure in a headline is not always money that actually moved.
| Deal | Amount | Nature of the amount |
|---|---|---|
| Pathos AI Series E | $237 million | Funds raised; led by a major US drugmaker |
| Pathos AI valuation | About $3 billion | Company valuation, not money invested |
| Warp Sciences launch | $215 million | Funds raised; says it will bring AI drug discovery to the lab |
| AI Proteins and a Japanese drugmaker | $400 million | Miniprotein discovery deal; unclear if firm or total value |
Amounts taken from the original reports. No currency conversion
The amounts listed are very different in kind. Some is cash actually paid in. Some is a price investors put on a whole company. Some may include payments due only if conditions are met. A row of large sums makes a market look hot. But compare them without sorting their nature, and the story grows bigger than it is.
A price put on a company is not cash in its hands. The gap between that price and the cash paid in shows how much is expected later. Expectations shrink if results do not come. A high price is not an achievement.
When drug companies announce partnerships, the stated sum shapes how people react. Misread it, and you misjudge both the partner's strength and your own burden. Read the upfront payment apart from later payments tied to results. With firms that design medicines, results take years, so later payments tend to be large. Reporting only the headline sum inside the company misleads its readers.
Follow the money further and another stream appears. Prices are being put on data itself, not just on companies.
05Pharma's data deals
Source www.tokenpost.com ／ marketscreener.com ／ Indian Pharma Post
What a medicine-design model can do depends on what it reads. So the data itself has become something to trade.
Neurology data
Tempus AI and Pathos AI signed a neurology data deal worth up to $55 million. That is a ceiling, not a fixed amount.
Licensing oncology data
Tempus AI signed a multi-year deal to license multimodal oncology data to a major US drugmaker. No amount was reported.
From ECGs to AI care
Tricog Health was reported to expand globally with a European drugmaker and OMRON Healthcare. No amount in the headline.
Most of these deals are about getting data. One caps the amount, and others keep it hidden. More striking than any sum is the focus on narrow disease areas, collected in depth. The partners include drug makers. They may value owning what feeds the model more than owning its maker. Note also that one name shows up on both the investment side and the data side. I read that as locking up money and data together.
Lay images over test results, and links invisible in one record start to show. So data collected as a set is worth more. What counts is the range of combinations, more than sheer volume.
Drug companies hold many records from trials and post-launch studies. Those records may be used only within what patients agreed to. Whether you sell or buy, check consent and anonymization first. Knowing what your records are worth helps you protect them and put them to use.
However much data and money pile up, a person makes the final call. In the last chapter, we see what AI changed in expert work.
06AI and expert judgment
Source The Decoder ／ World Trademark Review ／ Washington Times
The faster AI makes the work, the more human judgment stands out.
Junior patent attorneys
A study found AI quickly lifts junior patent attorneys' output but may hinder their professional growth (The Decoder).
Trademark search and legal calls
A study warned that AI is strong at trademark searches but weak at legal judgment (World Trademark Review).
A federal judge who used AI
A federal judge who used AI was reported removed from a DEI-related case (Washington Times).
Every reported case involves expert work. AI lifts the performance of beginners within a short time, and it is good at searching. Yet one study warns it is poor at judgments made under the law. And a person whose role is to judge stepped away from a case after relying on it. People will need to explain afterward whether they relied on AI.
What worries me is the lost chance to grow. Young people build judgment by searching, hesitating, and being corrected by seniors. If AI takes over that process, today's work improves. But in some years, nobody may be left who can judge. Still, this is a concern raised by research, not a proven result.
In pharma too, reviewing promotional materials and assessing safety end with a person taking responsibility. If you use AI, decide first what it handles and where people take over. Leaving room for juniors to judge on their own matters just as much. Record who checked each answer.
Deciding how much to hand to AI is the same kind of work as checking where things come from. Now, back to today's question.
Watch whether Alphabet, Microsoft and Amazon, reporting on October 28–29, show their capex scale and how AI revenue pays it back.
Open the full transcript
Intro
Is an AI worth choosing only because it is smart?The answer is no. This video shows what else decides the choice. People who own the rights to songs have taken an AI developer to court. On the same day, a national agency was reported to have narrowed the AI it allows. When drug makers choose AI, the same question comes back to them. What did the system learn from?
Which path did its answer take?Whose money keeps it running?Being able to explain that is starting to decide which AI gets picked. Let us follow the day's events in that order.
CH 01 Music copyright suits against AI
We begin with songs. What an AI learned from is now being tested in court.The ones suing hold the rights to the music. From the headlines, their point is that songs were fed to a model without consent. Until now, what a developer learned from stayed an internal matter. Now it must be explained to outsiders. I think the plaintiffs want more than money. They want to know how their work was used.The dispute is not about the answers a model gives. It is about an earlier step. Was consent obtained before learning began?
The sum looks heavy for one company. That pressure will push other developers to rethink how they gather material, and to record where it came from.Whether that material was gathered properly will become one test when we judge an AI.Pharma is not exempt. If you train an in-house model on internal documents or papers, the same question returns to you. Confirm you hold the rights before you start. Are your own files being read by outside systems without your knowledge?
If the material is in question, the way users choose AI changes too. Next, a nation acting as a user.
CH 02 Governments start choosing AI
On the same day the makers were held to account, users moved too. Government offices began deciding for themselves which AI they may use.One country's agency stopped popular chat assistants. It kept only the assistant built into its office software. I doubt the survivor won on smarts. It sits inside systems the agency already runs, so it is easier to control. The reports do not give a reason. My guess is that the deciding point was whether the agency could watch where its information goes. Elsewhere, local governments are pressing for national rules.Inviting developers into the research suggests a wish to curb misuse while models are still being built. States are no longer only picking tools. They are stepping in between makers and users.In a drug company, if staff each use any AI they like, patient data or unpublished trial results could leak. So the company should name the AI it allows, and record why. Keep someone in charge of reviewing the usage rules after the choice, so decisions stay steady. But what if another AI was answering behind the one your company chose?
Next, where requested work is really processed.
CH 03 AI that routes the work
The AI you talk to on screen may no longer be the one producing the answer.Reports say work sent to one company's assistant is quietly passed to another company's model. To the user, nothing seems to change. Yet a different firm writes the answer. The report argues that the advantage has moved from raw intelligence to deciding where work goes. If so, ranking models by smartness will matter less and less.Price will weigh heavily in that choice. The two costs compared differ by more than an order of magnitude. With a gap like that, a front desk will want to send work to the cheaper side. The problem is that users can neither know nor choose who answered.For pharma, the danger is losing track of which company's system handled your documents. With patient information, you must be able to say which firm, in which country, processed it. Before signing, ask the vendor to disclose where work goes, and to warn you before changing it. Choosing on price alone pushes that check aside. While routing and price unsettle what an AI is worth, big money is flowing to firms that design medicines with AI. Next, what those sums really mean.
CH 04 Money into drug-discovery AI
A big figure in a headline is not always money that actually moved.The amounts listed are very different in kind. Some is cash actually paid in. Some is a price investors put on a whole company. Some may include payments due only if conditions are met. A row of large sums makes a market look hot. But compare them without sorting their nature, and the story grows bigger than it is.A price put on a company is not cash in its hands. The gap between that price and the cash paid in shows how much is expected later. Expectations shrink if results do not come. A high price is not an achievement.When drug companies announce partnerships, the stated sum shapes how people react. Misread it, and you misjudge both the partner's strength and your own burden. Read the upfront payment apart from later payments tied to results. With firms that design medicines, results take years, so later payments tend to be large. Reporting only the headline sum inside the company misleads its readers. Follow the money further and another stream appears. Prices are being put on data itself, not just on companies.
CH 05 Pharma's data deals
What a medicine-design model can do depends on what it reads. So the data itself has become something to trade.Most of these deals are about getting data. One caps the amount, and others keep it hidden. More striking than any sum is the focus on narrow disease areas, collected in depth. The partners include drug makers. They may value owning what feeds the model more than owning its maker. Note also that one name shows up on both the investment side and the data side. I read that as locking up money and data together.Lay images over test results, and links invisible in one record start to show. So data collected as a set is worth more. What counts is the range of combinations, more than sheer volume.Drug companies hold many records from trials and post-launch studies. Those records may be used only within what patients agreed to. Whether you sell or buy, check consent and anonymization first. Knowing what your records are worth helps you protect them and put them to use. However much data and money pile up, a person makes the final call. In the last chapter, we see what AI changed in expert work.
CH 06 AI and expert judgment
The faster AI makes the work, the more human judgment stands out.Every reported case involves expert work. AI lifts the performance of beginners within a short time, and it is good at searching. Yet one study warns it is poor at judgments made under the law. And a person whose role is to judge stepped away from a case after relying on it. People will need to explain afterward whether they relied on AI.What worries me is the lost chance to grow. Young people build judgment by searching, hesitating, and being corrected by seniors. If AI takes over that process, today's work improves. But in some years, nobody may be left who can judge. Still, this is a concern raised by research, not a proven result.In pharma too, reviewing promotional materials and assessing safety end with a person taking responsibility. If you use AI, decide first what it handles and where people take over. Leaving room for juniors to judge on their own matters just as much. Record who checked each answer. Deciding how much to hand to AI is the same kind of work as checking where things come from. Now, back to today's question.
Wrap-up
An AI's worth was not decided by smartness alone. What did it learn from, which path did its answer take, whose money runs it, and who judges in the end?Only AI whose origins can be explained will stay in use in pharma, I believe. Soon, large tech firms report results. Will they show how they earn back what they spent?
I will follow that money.
- CH 01Billboard「Anthropic Sued For Stealing Elvis Presley Music to Train AI: ‘Deeply Protective of These Iconic Songs’」 billboard.com
- CH 01Law Commentary「Suno and Anthropic Face Nearly $340 Million in Copyright Claims Over Songs by Madonna, Johnny Cash」 lawcommentary.com
- CH 02Yahoo「Sweden's Government Offices Block ChatGPT and Claude, Keep Microsoft Copilot」 yahoo.com
- CH 02Unite.AI「UK Establishes AI Biosecurity Lab Backed by OpenAI and Anthropic」 unite.ai
- CH 02The Daily Tar Heel「North Carolina joins 25 states in urging Congress for AI regulation」 dailytarheel.com
- CH 03Forkast News「Grok Bot Will Route Your Tasks to Claude. The Moat Just Moved to the Switchboard.」 forkast.news
- CH 03https://tech-insider.org/「DeepSeek vs Claude: $0.15 vs $4 Per 1M Tokens [2026]」 tech-insider.org
- CH 04Wowtale「Pathos AI Raises $237M Series E at ~$3B Valuation, Led by Bristol Myers Squibb」 en.wowtale.net
- CH 04Kingy AI「Warp Sciences Launches With $215 Million to Bring AI Drug Discovery Into the Lab」 kingy.ai
- CH 04BioSpectrum Asia「Japan's Astellas Pharma inks $400 M deal with AI Proteins to develop novel miniprotein-based therapeutics」 biospectrumasia.com
- CH 05www.tokenpost.com「Tempus AI and Pathos AI Sign Neurology Data Deal Worth Up to $55 Million」 tokenpost.com
- CH 05marketscreener.com「Tempus AI, Inc. Enters Into Multi-Year Agreement With Bristol Myers Squibb To License Multimodal Oncology Data」 marketscreener.com
- CH 05Indian Pharma Post「From ECGs to AI-powered Care: Tricog Health scales global ambitions with Sanofi & OMRON Healthcare」 indianpharmapost.com
- CH 06The Decoder「AI makes junior patent attorneys instantly better but may quietly stunt their professional growth」 the-decoder.com
- CH 06World Trademark Review「AI excels at trademark clearance but falls short on legal judgement, study warns」 worldtrademarkreview.com
- CH 06Washington Times「Federal judge who used AI gets booted off DEI case」 washingtontimes.com
Articles used
Today's related reports
- AI Daily News October 25, 2026
- AI & Economy News October 25, 2026
- AI & Finance News October 25, 2026
The narration is synthetic speech. The content draws only on this site's articles from the same day. Each edition passes seven plain-language checks before publication. That means known obstacles to comprehension are held below threshold — it is not a guarantee of comprehension.