Today’s three AI news stories:
One says:
Don’t give up the leading position because of fear of AI.
Another says:
Open source AI is not just a company strategy; it can become a joint effort among an international group.
The third says:
Core chips and strategic technologies are so important that criminal laws must be amended accordingly.
Putting these three together:
We see that AI competition is changing.
In the past, the common questions were:
Who is stronger, OpenAI or Anthropic?
How different are US and Chinese AI models?
Which benchmark ranks first?
Now countries are asking:
Which capabilities must accelerate?
Which technologies should be shared?
What must not be leaked?
AI has gradually shifted from a tech product to:
National capability.
First story: Trump publicly opposes large-scale AI slowdowns over risks
Recently, the AI industry showed a rare scene.
Anthropic CEO Dario Amodei:
Called on frontier AI companies to consider slowing the growth of model capabilities.
Sam Altman also indicated OpenAI’s willingness to discuss this.
Elon Musk likewise supported some of these proposals.
The reason:
Model capabilities are advancing rapidly.
Safety mechanisms, monitoring, and regulations might not keep pace.
But US President Trump, on September 13, delivered a completely different political answer.
Trump believes some AI fears are exaggerated
In an interview in Ireland, Trump said:
Some people are exaggerating certain AI risks.
He described these voices as:
Very negative forces.
His core concern was straightforward:
The US currently leads China in AI.
If the US slams the brakes hard out of fear of potential future risks,
The lead may be lost.
This does not mean Trump says “No need to regulate AI”
This distinction is important.
Trump did not say all AI regulations should be repealed.
His position is closer to this:
Certain guardrails can exist,
But very distant, highly uncertain, or catastrophic assumptions cannot be allowed to stop the entire US AI industry’s progress.
In other words:
Safety discussions are necessary.
But:
We must not lose the competitive edge.
This stands in sharp contrast with recent AI lab sentiments
Anthropic recently advocated:
The pace of capability growth needs to allow safety work to catch up.
OpenAI also started saying:
The major AI labs can discuss slowing if they coordinate.
The real difference between these positions isn’t just belief in AI risks, but:
Who bears the strategic cost of “slowing down first”?
If only one US company slows and others don’t?
That company risks losing model leadership, enterprise customers, talent,
and even capital.
If all US companies slow down, but Chinese firms don’t,
The issue turns into:
National security.
Thus, AI slowdowns will be difficult to decide by tech companies alone.
They directly confront:
International competition.
Trump’s real argument is about “relative capability”
Assume:
The US AI capability is 100.
China’s is 90.
Both advance, but the US stops at 100 while China continues upward from 90.
This is politically a very different situation.
So governments ask not just:
Is AI dangerous?
But also:
Will others’ AI become stronger than mine?
That is why AI safety and national security increasingly overlap.
AI companies may want a joint slowdown but governments may disagree
If OpenAI, Anthropic, and xAI agree to limit capability growth to a certain pace in a year,
Theoretically, safety research gets more time.
But if governments think China is rapidly catching up,
They may ask:
Why are you limiting US capabilities yourselves?
This creates conflict between:
Corporate safety governance and national strategy.
Thus, the real challenge in AI pacing is not agreeing on danger
But whether all parties can slow down simultaneously.
If not, any side slowing alone feels disadvantaged.
This is very similar to:
Arms control – nuclear weapons, missiles, or other strategic high-tech.
Everyone knows unrestrained competition is risky,
But no one wants to disarm first.
Second story: Xi Jinping proposes BRICS AI open source community
On the same day, a completely different AI strategy emerged.
At the 18th BRICS summit held in New Delhi, India, Xi Jinping put forward five cooperation initiatives.
The first focused directly on:
AI.
Not by building new data centers or buying more GPUs,
But by:
Establishing a BRICS AI Open Source Community.
What does China want BRICS to do together?
Xi’s proposals include:
Building a BRICS AI open source community.
Supporting joint development and application of large language models.
Organizing specialized AI seminars and training courses.
Creating a more open AI ecosystem.
In other words:
China wants to make open source AI an official entry point for BRICS technology cooperation.
BRICS is no longer just the original five countries
Many still think BRICS means:
Brazil, Russia, India, China, and South Africa.
But it has been expanding to include more emerging markets and Global South countries.
If open source AI cooperation truly starts building shared:
Models, training, applications,
and infrastructure,
Then the impact is more than just:
Another GitHub project.
It could become:
Another AI technology ecosystem.
Why does China emphasize open source?
Because not every country can train frontier models on its own.
Developing a frontier AI lab requires:
GPUs, talent, data, power, cloud resources, funds, and research ability.
Many developing countries don’t have all these at once.
Open source offers an alternative:
No need to start from zero.
You can directly:
Take a model, deploy it locally, fine-tune it, change the language, and create local applications.
This is also a key direction of China’s AI competition
The most influential US frontier models
are often closed and accessed via API, cloud, or subscription.
China has many companies producing open-weight, open-source,
and easier-to-deploy local models.
So the competition is not just about:
Whose model is best?
But also:
Who can make their AI accessible to more countries?
For the Global South, this is a very practical issue
Suppose a country wants to build AI for government, education, and healthcare in local languages.
It may not want to send all data to US clouds,
Or have a large dollar budget to buy expensive APIs long-term.
If models can run locally, be modified, and trained independently,
Their appeal naturally grows.
In that sense, open source is also a form of:
Technology diplomacy.
Open source AI ecosystems also bring standards influence
If many countries use the same model family, developer tools, dataset standards, and deployment stacks,
Over time, what’s built is not just a model,
But an ecosystem.
Like Android, Linux, and cloud platforms,
The hardest thing to replace is not a version but the number of users.
So China is really competing for the “AI entry point”
Many emerging markets building their national AI systems for the first time will ask:
Which model, cloud, framework, and open source standard to use?
If early on they enter a China-led BRICS open source ecosystem,
Future talent, training, models, infrastructure, and partnerships are likely to form around it.
This is ecosystem competition.
But “open source” does not mean no national interests
Open source lowers barriers, promotes research, and boosts transparency,
But countries supporting open source can still have strategic goals, such as:
Expanding model adoption, developer communities, technical standards, and international clout.
So open source and geopolitics are not contradictory.
They can even become tools for geopolitics.
Third story: South Korea’s updated espionage law goes into effect
While the US debates speeding up AI,
China expands its AI ecosystem,
South Korea addresses another issue:
How to stop core technology leaks.
On September 13, a new criminal law took effect in South Korea.
One major change is expanding the scope of espionage crimes.
South Korea’s espionage laws have a heavy historical burden
Its unique national security environment long focused espionage laws on North Korea,
which is its enemy state.
But with tech industry developments, new problems emerged.
If someone passes sensitive info on semiconductors, batteries, displays, or advanced manufacturing to other countries,
Traditional espionage law may not apply.
The new law now covers espionage for “foreign countries or similar organizations”
South Korea’s new Criminal Act adds new crime types beyond espionage for “enemy states.”
It now includes:
Espionage on behalf of foreign countries or similar groups
Whereby, under instruction, demand, or contact, people collect, leak, transmit, or mediate national secrets.
Punishable by a minimum of three years imprisonment.
Why is this especially important now?
Because South Korea holds some of the world’s most vital AI physical capabilities.
Samsung and SK Hynix,
Especially HBM (High Bandwidth Memory),
Which are critical parts of AI accelerators, servers, and large-scale model training.
Today, AI companies want more computing power, not just GPUs, but also HBM, packaging, and advanced semiconductor manufacturing.
What was once a trade secret increasingly is seen as a:
National strategic asset.
South Korea has faced multiple tech leakage cases in recent years
Semiconductor engineers have switched jobs,
Taking production documents, designs, equipment data, and factory information.
This made industry insiders see past laws as insufficient for dealing with cases that are not North Korean espionage, but foreign tech acquisition.
The updated criminal law fills this gap.
But the new espionage law does not mean all company data leaks are espionage
It’s important to clarify.
The law focuses on national secrets and requires elements like instructions or communications by foreign or similar organizations.
Normal business disputes, regular employee moves or trade secret cases are not automatically espionage.
South Korea already has other industrial technology protection laws.
This change is not:
“An employee taking a company document automatically becomes a spy,”
But rather:
The state is expanding the criminal framework for foreign technology infiltration.
How is this directly related to AI?
AI competition ultimately depends on the physical supply chain.
Models need chips, memory, packaging, networking, power, and data centers.
If a country’s model isn’t the world’s best but it controls irreplaceable HBM,
It still holds huge AI strategic value.
So national AI competition is not just OpenAI vs. DeepSeek,
But also whether SK Hynix’s process can be protected.
These three news stories show three opposite approaches
The US says:
Don’t slow down.
China says:
Expand sharing.
South Korea says:
Strengthen protection.
But the underlying logic is the same:
Every country views AI as:
National competitiveness.
They just have different advantages and thus different strategies.
The US’s biggest advantage is frontier models and computing ecosystems
It fears losing speed advantage by self-imposed limits.
Trump’s political instinct is:
Don’t lose leadership out of fear.
China has many models, engineering talent, and strong Global South ties
It uses open source, training, and infrastructure to:
Spread its AI ecosystem to more countries.
The strategy is:
Get more countries into its tech network.
South Korea controls world-class semiconductor tech
Its focus:
Is not to build another ChatGPT,
But to protect:
Memory, manufacturing capabilities, and process know-how.
Its strategy naturally becomes:
Preventing core tech from leaking.
This is the most interesting aspect of AI entering national competition
The same technology gets totally different conclusions by different countries.
Model leaders want to go fast.
Ecosystem expanders want openness.
Key suppliers want protection.
Hence, future AI debates can’t be resolved by statements like:
“All countries should support open AI.”
Or “AI should be fully regulated.”
Each layer has different interests.
“Open source” and “protection” can coexist
China can open some AI models,
While restricting exports of certain models, chips, and technologies.
South Korea can encourage global chip trade,
While strictly protecting process know-how.
The US can promote global AI developer ecosystems,
While limiting advanced chip exports to some regions.
This is not totally contradictory.
Countries usually want:
Others to use their technology,
But not to take their core capabilities away.
AI is becoming like other strategic industries
Aviation, nuclear, communications, chips, defense—all share a characteristic.
Governments don’t just say, “Let the market decide.”
Because these fields involve economy, security, diplomacy, supply chains, and autonomy.
AI now officially joins this category.
What impact does this have on ordinary people?
It may be more direct than imagined.
First:
Which AI models you can use may be affected by national policies, export regulations, and regional restrictions.
Second:
AI service pricing will be influenced by GPUs, HBM, power, and supply chains.
Third:
Enterprise AI use will face questions on whether data can leave the country, cross-border model use, and which tech can be shared with overseas branches and partners.
For businesses, “Which AI to use” may no longer be just an IT choice
Previously, choosing SaaS compared price, functionality, and interface.
Now, AI choices might also consider:
Model origin, data location, open or closed source, export restrictions, supply chain risks, and security policies.
Future large company AI purchasing might look more like buying critical infrastructure than just another productivity app.
AI globalization is becoming two simultaneous forces
One force is:
AI must involve cross-border cooperation,
Because safety, standards, and research require joint efforts.
The other force is:
AI must be nationalized,
Because models, chips, data, and supply chains increasingly relate to national strategy.
These forces coexist.
So we see the US and China competing while discussing AI safety,
China promoting open source while protecting core tech,
South Korea exporting chips globally while strengthening penalties on tech leaks.
This is not contradiction,
But AI’s dual nature of being global and national.
The key takeaway: AI’s unit of competition is getting larger
Previously, AI competition units were:
Models.
Then companies.
Now it is:
Countries.
When countries join, comparisons cover not only benchmarks,
But research talent, computing power, chips, energy, laws, capital, supply chains, allies, and open source ecosystems.
So next time you see AI news,
Besides asking:
“How much stronger is the model now?”
You can also ask:
“What part of AI is this country aiming to make its long-term advantage?”
This may better explain the next phase of AI competition than any single benchmark.
Today, progress a little with AI,
Learn one AI skill every day,
Save a bit of time daily,
Improve a little every day.
SasaDaily, growing with you.
