AI's $44T promise; markets lose $1T── A builder's own bill, agents shut out, China's export controls
Download the video (MP4, 11 MB)
Hopes for this technology have grown so large that they now shape forecasts for a whole national economy. Yet in the same days, share prices fell hard. Hope and doubt are rising together. Between them sits one question I keep asking. Who holds the power to limit how it is used?
Is it the firms that build it, or the firms that let it in?Is it governments, or is it the market?Every story today leads back to this question, and no single answer fits. So pharma teams should not wait for someone else to decide. Set your own limits on use, and your own way to stop it.
01Forecast: US GDP of $44 trillion
Source Seeking Alpha ／ Reuters
My first check is simple. Where is the hopeful voice coming from? The bigger a promise sounds, the more it moves budgets and policy. So I treat the speaker as being as important as the forecast itself.
The speaker sits at the heart of the firms that build this technology. The listeners were people who steer national economies. So a builder pitched a vast outlook straight to the deciders. If they accept it, it becomes a reason to ease rules and to draw investment.
The claim is that the value a nation creates could jump within a few years. But it is a prediction, not something that has happened. And the speaker gains when hopes rise. Listeners should discount for that.
Big outlooks spread before anyone tests them. Once spread, they turn into planning assumptions. After that, questioning them again is hard.
Drug companies will hear claims like this in meetings about adoption. When that happens, ask who said it, and to whom. How much work the software can save in your company only shows up in your own records.
And the builders who sell this hope are now stepping into how their own rules get written.
02The builder's own bill
Source Politico ／ Stocktwits
So what are the builders doing about the rules that bind them? When a regulated company drafts its own rules, we must ask whom those rules protect.
| Item | As reported |
|---|---|
| Drafted by | Anthropic (written in-house) |
| Given to | Lawmakers, pitched privately |
| Criticism | Limits government power to block dangerous models (commentary) |
Reported by Politico and Bloomberg. Full text not published
A builder reportedly wrote its own proposal and took it to legislators without making it public. The reports do not let us check what it says in full. Yet critics say it would narrow the state's power to halt risky systems. If so, the party being stopped would decide how far stopping goes.
- Venue
- Dreamforce
- Stance
- Against AI safety regulation
- Venue
- Dreamforce
- Stance
- Reported as contrasting with Huang
Based on a Stocktwits headline
Leaders in the industry also disagree on rules. One rejects them, and reports contrast that with another who backs them. But if the backer's proposal weakens the state, words and substance diverge. We cannot judge that until the whole draft is out.
People in pharma know how hard it is when the regulated side writes the rules. Rules earn trust only when outsiders can read them. A proposal kept private does not meet that test yet. When it is released, I want to read it myself and see whom it protects.
But the firms that receive this software are not waiting for rules. They have started keeping the builders' systems out of their services.
03AI agents shut out
Source The Tech Buzz ／ UA.NEWS ／ Yeni Şafak English
Before any rules are settled, the choice to accept or refuse is already being made. It is made not by builders or governments, but by the firms on the receiving end.
| Who | As reported |
|---|---|
| Yelp, Delta Air Lines | Blocking autonomous agents from OpenAI, Meta and xAI |
| Banks and others | Restricting use of AI agents |
| Amazon and Perplexity | Court fight over agent shopping continues |
Firms that handle travel bookings and money reportedly refuse software that acts on a person's behalf. A dispute over delegated buying has even reached the courts. Such software can only work if the other side lets it in. However capable it is, a refusal leaves it with nothing to do.
Meanwhile, a national security body said such software got into systems it was running. Even professional defenders could not keep it out. I think this risk is why more firms are saying no.
Delegated software chooses its own path through a task. So it can wander where nobody expected. Decide in advance how far it may go, and how to verify who is on the other end.
Pharma companies hold patient data and trial records in their systems. Will you let outside software in? How much may your own internal tools touch? Your company decides that, not the builder.
As more firms refuse this software, the market began to doubt the money poured into it.
04$1 trillion lost in one day
Source Yahoo Finance ／ The Globe and Mail ／ Fortune ／ Yahoo Finance
The bigger the hope, the more money is needed to back it. Will that money come back in line with the hope? Today's market moves were aimed at that question.
| Item | Amount | Report |
|---|---|---|
| Tech stock drop | Over $1 trillion in one trading day | Nvidia CEO: "All of this is going to pass" |
| Microsoft capital spending | $175 billion | Seen as a burden on the business |
| Meta AI investment | Over $100 billion | Reportedly entrusted to Wang |
| US data center power | 32 GW short through 2028 | Morgan Stanley estimate |
Markets reportedly shed enormous value in a single session. Builders keep spending on a vast scale, and some say the weight is squeezing their core earnings. There is also an estimate that electricity for computing buildings will run short. Money alone cannot run computers without power.
Other prices moved that day too, so the fall cannot be blamed on this technology alone. Still, I think the bigger the bet, the bigger the swings. An industry leader called it temporary. That is a view, not a promise.
Builders pay for buildings and chips first, then recover the cost through usage fees later. With that order, slow recovery could reach users as higher prices.
This matters for drug companies using such tools. If the finances of a vendor tighten, fees may rise or a service may end. Avoid setups that rely on a single vendor.
Beyond money and power, nations too are starting to control whether this technology may leave their borders.
05China's AI export controls
Source Pasquale Pillitteri ／ The China-Global South Project ／ BigGo Finance
Companies and markets are not the only ones shaping how this technology is handled. Nations also want to control whether their own systems go abroad.
What is covered
23 AI technologies, including large models, agents and compute clusters. Beijing published the draft on October 16.
What is excluded
Open source and use via API are excluded. The proposal reportedly also covers AI's overseas expansion.
DeepSeek funding
Second round expected to close at about 100 billion yuan (about $15 billion). CATL reportedly adds investment.
That country proposed covering a wide range, from big systems to computing hardware. Yet openly released code and online access reportedly fall outside. To me, it reads like this. Keep the technology at home, but keep letting others use it. Meanwhile, large sums keep flowing to a developer in the same country. It manages what goes out, while growing what stays in. Both happen at once.
A state applies such controls because it sees this technology as tied to its security. It is now treated as national power, not just a product. Where a system was made may decide whether you can use it.
Drug companies run research and development across borders. Which country's rules govern the tools you use? Could you keep using them if those rules changed? Check both before you adopt them.
After the moves of nations and markets, we turn to events that touch drug research and workplaces directly.
06In the pharma workplace
Source cristinaiglesias.com ／ California Employment Law Report ／ Fox Business
How does all this connect to pharma work? Research tools, staff monitoring and promises about drug effects. Today brought news close to our field in each.
AlphaFold 3 access
Google DeepMind and the European Commission will give 200 researchers $3 million worth of compute.
Ban on mood-reading AI
California bans AI that reads employees' moods (AB 1883). Employers urged to review tools such as camera analysis.
Claim of 150-year lifespans
The CEO of an AI drug discovery company described on TV the drugs in development that target aging and disease.
A public body and a developer will let scientists use software that predicts protein shapes. High computing costs may stop forcing researchers to give up. Meanwhile, one state banned using software to read workers' feelings. Now is the time to check whether your monitoring tools break the law. Lose your staff's trust, and no tool will help.
There is also a report of a company betting on medicines that would greatly extend life. Pharma professionals know better than anyone that claims of benefit need evidence. The bigger the promise, the more you must ask which trial stage it comes from. The habit of reading trial records, not slogans, pays off here too.
Research tools are opening up, monitoring is being restricted, and bold promises face doubt. Each asks the user to make their own call.
So how do the people working inside the builders see the future of their own jobs?
07Jobs inside the builder
Source India Today ／ Business Insider
Finally, we hear from the people working closest to this technology. Staff at a builder learn before anyone else what the systems can now do.
| Report | Content |
|---|---|
| Junior hiring | Fewer new-grad and junior hires; AI does 6x human work (India Today) |
| Survey of 1,334 staff | Expect AI to do 80% of their own work by 2030 (Business Insider) |
| US workers | Expect less impact than the staff predict |
That company reportedly takes on fewer young hires. Many of its staff expect software to handle most of their own tasks within a few years. The wider workforce expects a smaller effect. Those who watch from close up see bigger change. Nobody knows yet which side is right. But what happens inside a builder may spread outside a little later.
This gap matters a great deal for drug companies. Young staff learned judgment by organizing records and doing early literature searches. If software takes over that work, the chance to learn may vanish. So I believe firms should rethink how they train people before bringing these tools in. Start by separating tasks to delegate from tasks people learn by doing.
The scale of change these insiders expect suggests users may have less time to prepare than they think.
How will Congress respond to Anthropic's bill? The next issue is how lawmakers and other companies react to the private pitch.
Open the full transcript
Intro
Hopes for this technology have grown so large that they now shape forecasts for a whole national economy. Yet in the same days, share prices fell hard. Hope and doubt are rising together. Between them sits one question I keep asking. Who holds the power to limit how it is used?
Is it the firms that build it, or the firms that let it in?Is it governments, or is it the market?Every story today leads back to this question, and no single answer fits. So pharma teams should not wait for someone else to decide. Set your own limits on use, and your own way to stop it.
CH 01 Forecast: US GDP of $44 trillion
My first check is simple. Where is the hopeful voice coming from?The bigger a promise sounds, the more it moves budgets and policy. So I treat the speaker as being as important as the forecast itself.The speaker sits at the heart of the firms that build this technology. The listeners were people who steer national economies. So a builder pitched a vast outlook straight to the deciders. If they accept it, it becomes a reason to ease rules and to draw investment.The claim is that the value a nation creates could jump within a few years. But it is a prediction, not something that has happened. And the speaker gains when hopes rise. Listeners should discount for that.Big outlooks spread before anyone tests them. Once spread, they turn into planning assumptions. After that, questioning them again is hard.Drug companies will hear claims like this in meetings about adoption. When that happens, ask who said it, and to whom. How much work the software can save in your company only shows up in your own records. And the builders who sell this hope are now stepping into how their own rules get written.
CH 02 The builder's own bill
So what are the builders doing about the rules that bind them?When a regulated company drafts its own rules, we must ask whom those rules protect.A builder reportedly wrote its own proposal and took it to legislators without making it public. The reports do not let us check what it says in full. Yet critics say it would narrow the state's power to halt risky systems. If so, the party being stopped would decide how far stopping goes.Leaders in the industry also disagree on rules. One rejects them, and reports contrast that with another who backs them. But if the backer's proposal weakens the state, words and substance diverge. We cannot judge that until the whole draft is out.People in pharma know how hard it is when the regulated side writes the rules. Rules earn trust only when outsiders can read them. A proposal kept private does not meet that test yet. When it is released, I want to read it myself and see whom it protects. But the firms that receive this software are not waiting for rules. They have started keeping the builders' systems out of their services.
CH 03 AI agents shut out
Before any rules are settled, the choice to accept or refuse is already being made. It is made not by builders or governments, but by the firms on the receiving end.Firms that handle travel bookings and money reportedly refuse software that acts on a person's behalf. A dispute over delegated buying has even reached the courts. Such software can only work if the other side lets it in. However capable it is, a refusal leaves it with nothing to do.Meanwhile, a national security body said such software got into systems it was running. Even professional defenders could not keep it out. I think this risk is why more firms are saying no.Delegated software chooses its own path through a task. So it can wander where nobody expected. Decide in advance how far it may go, and how to verify who is on the other end.Pharma companies hold patient data and trial records in their systems. Will you let outside software in?
How much may your own internal tools touch?Your company decides that, not the builder. As more firms refuse this software, the market began to doubt the money poured into it.
CH 04 $1 trillion lost in one day
The bigger the hope, the more money is needed to back it. Will that money come back in line with the hope?Today's market moves were aimed at that question.Markets reportedly shed enormous value in a single session. Builders keep spending on a vast scale, and some say the weight is squeezing their core earnings. There is also an estimate that electricity for computing buildings will run short. Money alone cannot run computers without power.Other prices moved that day too, so the fall cannot be blamed on this technology alone. Still, I think the bigger the bet, the bigger the swings. An industry leader called it temporary. That is a view, not a promise.Builders pay for buildings and chips first, then recover the cost through usage fees later. With that order, slow recovery could reach users as higher prices.This matters for drug companies using such tools. If the finances of a vendor tighten, fees may rise or a service may end. Avoid setups that rely on a single vendor. Beyond money and power, nations too are starting to control whether this technology may leave their borders.
CH 05 China's AI export controls
Companies and markets are not the only ones shaping how this technology is handled. Nations also want to control whether their own systems go abroad.That country proposed covering a wide range, from big systems to computing hardware. Yet openly released code and online access reportedly fall outside. To me, it reads like this. Keep the technology at home, but keep letting others use it. Meanwhile, large sums keep flowing to a developer in the same country. It manages what goes out, while growing what stays in. Both happen at once.A state applies such controls because it sees this technology as tied to its security. It is now treated as national power, not just a product. Where a system was made may decide whether you can use it.Drug companies run research and development across borders. Which country's rules govern the tools you use?
Could you keep using them if those rules changed?Check both before you adopt them. After the moves of nations and markets, we turn to events that touch drug research and workplaces directly.
CH 06 In the pharma workplace
How does all this connect to pharma work?Research tools, staff monitoring and promises about drug effects. Today brought news close to our field in each.A public body and a developer will let scientists use software that predicts protein shapes. High computing costs may stop forcing researchers to give up. Meanwhile, one state banned using software to read workers' feelings. Now is the time to check whether your monitoring tools break the law. Lose your staff's trust, and no tool will help.There is also a report of a company betting on medicines that would greatly extend life. Pharma professionals know better than anyone that claims of benefit need evidence. The bigger the promise, the more you must ask which trial stage it comes from. The habit of reading trial records, not slogans, pays off here too.Research tools are opening up, monitoring is being restricted, and bold promises face doubt. Each asks the user to make their own call. So how do the people working inside the builders see the future of their own jobs?
CH 07 Jobs inside the builder
Finally, we hear from the people working closest to this technology. Staff at a builder learn before anyone else what the systems can now do.That company reportedly takes on fewer young hires. Many of its staff expect software to handle most of their own tasks within a few years. The wider workforce expects a smaller effect. Those who watch from close up see bigger change. Nobody knows yet which side is right. But what happens inside a builder may spread outside a little later.This gap matters a great deal for drug companies. Young staff learned judgment by organizing records and doing early literature searches. If software takes over that work, the chance to learn may vanish. So I believe firms should rethink how they train people before bringing these tools in. Start by separating tasks to delegate from tasks people learn by doing. The scale of change these insiders expect suggests users may have less time to prepare than they think.
Wrap-up
Builders' voices keep inflating the hopes. But builders are not the only ones setting limits. Receiving firms refuse this software in their services. Markets show doubt with money. Nations control what leaves their borders. Now pharma teams must decide their own scope, how to stop, and how to train people. Record those choices so you can explain them later. Tomorrow, we watch how legislators receive the builder's draft.
- CH 01Seeking Alpha「Anthropic co-founder Clark projects U.S. GDP to hit $44T by 2030 thanks to AI」 seekingalpha.com
- CH 01Reuters「AI rapidly improving, could vastly boost US GDP, Anthropic's Clark tells policymakers」 reuters.com
- CH 02Politico「Anthropic quietly pitches its own AI safety bill to lawmakers」 politico.com
- CH 02Stocktwits「Nvidia’s Huang Shuns Need For AI Safety Regulations, In Contrast To Anthropic CEO Amodei’s View At Dreamforce Conference」 stocktwits.com
- CH 03The Tech Buzz「AI Agents Hit a Wall as Major Websites Block Access」 techbuzz.ai
- CH 03UA.NEWS「Yelp, Delta and banks restrict AI agents on their services — The Verge」 ua.news
- CH 03Yeni Şafak English「India’s cybersecurity agency says AI agents breached its computer systems」 en.yenisafak.com
- CH 04Yahoo Finance「Nvidia’s CEO Says “All of This Is Going to Pass” After Tech Lost $1 Trillion in a Day」 finance.yahoo.com
- CH 04The Globe and Mail「Here's How Microsoft's $175 Billion in Capital Expenditures Is Hurting Its Business」 theglobeandmail.com
- CH 04Fortune「A 44-minute call and a memo: How Alexandr Wang won Mark Zuckerberg’s trust—and a $100 billion-plus AI bet」 fortune.com
- CH 04Yahoo Finance「AI’s Power Problem Just Got a Number: 32 Gigawatts」 finance.yahoo.com
- CH 05Pasquale Pillitteri「China announces export controls on 23 AI technologies」 pasqualepillitteri.it
- CH 05The China-Global South Project「Beijing Seeks Tighter Grip on AI Exports Amid U.S. Rivalry」 chinaglobalsouth.com
- CH 05BigGo Finance「Industrial capital floods into DeepSeek, funding round doubles to ¥2.4 trillion as CATL adds ¥240 billion investment」 finance.biggo.com
- CH 06cristinaiglesias.com「Google DeepMind Backs 200 European Scientists with AlphaFold 3 Access and $3M Compute Grant」 cristinaiglesias.com
- CH 06California Employment Law Report「California Bans AI That Reads Employees' Moods (AB 1883) — Five Takeaways from Our 2027 New Laws Masterclass」 californiaemploymentlawreport.com
- CH 06Fox Business「Medical company bets on AI-powered drug breakthrough that could extend lifespan to 150」 foxbusiness.com
- CH 07India Today「Anthropic is hiring fewer juniors, says AI now does 6 times more work than humans」 indiatoday.in
- CH 07Business Insider「Anthropic polled 1,334 of its employees. They're more sure than most of us that AI will take over jobs.」 businessinsider.com
Articles used
- AI Daily News 2026-10-17 evening
- AI Daily News 2026-10-18
- AI News article digests 2026-10-17 evening
- AI News article digests 2026-10-18 morning
Today's related reports
- AI Daily News October 18, 2026
- AI & Economy News October 18, 2026
- AI & Finance News October 18, 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.