One company now sits behind roughly one in every five units of data centre capacity in China. The Next Web reports that, according to research firm SemiAnalysis, ByteDance accounts for about one-fifth of the country's delivered capacity, and it leases nearly all of it.

Key takeaways

• China has more than 24 gigawatts of delivered data centre capacity, and ByteDance occupies about a fifth of it.

• ByteDance rents almost everything, which makes it the key customer for China's wholesale data centre operators.

• Combined capital spending by ByteDance, Alibaba, Tencent and Baidu is projected to reach US$100 billion in 2026.

What is wholesale data centre capacity?

Wholesale means a tenant leases a whole building or campus from an operator. Retail colocation means renting racks in a shared facility. AI workloads need large blocks of power in one place, so they favour wholesale.

The numbers

SemiAnalysis's new China Datacenter Model tracks more than 1,000 facilities run by over 60 companies. Its headline findings, as reported on 29 September 2026:

• China's delivered capacity exceeds 24GW. That's more than Europe, the Middle East and Africa combined, at roughly 14GW.

• Alibaba, Tencent and Baidu spent US$20 billion on capital expenditure in the second quarter alone, double the figure a year earlier. All three reported negative free cash flow for the quarter.

• With ByteDance included, the four companies spent about US$35 billion in 2024 and more than US$50 billion in 2025.

Rent, don't own

ByteDance's model is different from the American hyperscalers, which tend to build and own. By leasing, it can expand quickly without carrying every building on its own balance sheet.

The demand is reshaping the operators. US-listed GDS and VNET signed 1.3GW of wholesale orders in the first half of 2026, yet won only about a third of ByteDance and Alibaba's orders between 2024 and 2026. SemiAnalysis's summary is that the AI era is wholesale. Older retail sites near big cities are running at 30% to 50% utilisation and aren't suited to AI.

What to keep in mind

These are estimates from one research firm's model, not audited company disclosures. ByteDance is private and doesn't publish this data.

Why it matters

Chinese cloud and AI companies are major tenants in Southeast Asia's data centre markets, including Johor. How they buy capacity at home says a lot about how they'll buy it abroad: big wholesale blocks, leased, and fast. Operators and landowners in Malaysia should plan for that demand profile, and for the power and water it needs. Tools such as digital twins for low-carbon operations are part of how that load gets managed.

It also shows how much compute China is installing while it builds a domestic chip stack. DeepSeek and Huawei's open-source tools for Ascend processors are the software half of the same push.

Frequently asked questions

How much of China's data centre capacity does ByteDance use?

About one-fifth of more than 24GW, according to SemiAnalysis.

Does ByteDance own its data centres?

Mostly no. It leases nearly all of its capacity from wholesale operators.

How much are China's tech giants spending?

ByteDance, Alibaba, Tencent and Baidu are projected to spend about US$100 billion in 2026, up from over US$50 billion in 2025.

How does China compare with other regions?

Its delivered capacity of more than 24GW is larger than Europe, the Middle East and Africa combined, at roughly 14GW.

See also

• DeepSeek and Huawei open-source Ascend chip tools

• Nvidia expands AI deployments across Southeast Asia

• AI for smart cities and transportation

Sources

• The Next Web: ByteDance rents a fifth of China's data centre capacity, SemiAnalysis says

Nvidia's biggest advantage has never been only its chips. It's the software developers use to program them. On 30 September 2026, DeepSeek and Huawei went after that advantage by open-sourcing programming infrastructure for Huawei's Ascend AI processors, Open Source For You reports.

Key takeaways

• DeepSeek and Huawei released open-source compute and communication libraries for Ascend chips, plus an Ascend version of the TileLang programming language.

• The goal is to make it easier to move AI workloads onto Ascend hardware without Nvidia's CUDA.

• The two companies have also built a supernode architecture using 128 Ascend 950 chips.

What is CUDA, and why does it matter?

CUDA is Nvidia's software platform for programming its graphics processors. Most AI code is written for it, which is why switching to another chip is hard even when the hardware is good enough. Any rival needs its own developer tools before it can win real workloads.

What was released

• TileLang for Ascend: a high-level language that wraps Ascend's low-level C instructions. The report says it simplifies programming compared with lower-level approaches such as CUDA, while still letting developers tune performance.

• Computing and distributed communication libraries: these match versions DeepSeek previously released for Nvidia hardware, so existing workloads are easier to adapt.

Huawei separately maintains its Compute Architecture for Neural Networks, known as CANN, which supports projects such as PyTorch, Triton and vLLM.

Why it's significant

DeepSeek is one of China's best-known AI labs, so its support carries weight with developers. Tools that mirror what already exists for Nvidia lower the cost of trying Ascend. The report adds that DeepSeek and Huawei have optimised computation and communication across a 128-chip Ascend 950 supernode.

What to watch

Open-sourcing tools doesn't guarantee adoption. CUDA has years of libraries, tutorials and trained engineers behind it. The test is whether developers outside the two companies start using and contributing to the Ascend stack.

Why it matters

Chip supply is now a strategic question for the whole region. Malaysia is building data centres at pace, and buyers want more than one viable source of AI compute. A working open-source software stack for Ascend would give them bargaining power and a fallback. It also ties into the wider Chinese build-out, where ByteDance now rents about a fifth of the country's data centre capacity.

For the region's chip industry, the story sits alongside NUS's push into advanced packaging with Malaysia and Nvidia's own expansion in Southeast Asia. The supply chain is being redrawn from several directions at once.

Frequently asked questions

What did DeepSeek and Huawei release?

Open-source programming infrastructure for Huawei's Ascend AI processors: compute and communication libraries, and an Ascend version of TileLang.

What is TileLang?

A high-level programming language for AI accelerators. The Ascend version wraps the chip's low-level C instructions to make development simpler.

Does this replace Nvidia's CUDA?

Not yet. It offers an alternative for Ascend hardware, but CUDA still has a much larger ecosystem.

What is the Ascend 950?

One of Huawei's AI processors. DeepSeek and Huawei have built a supernode architecture that uses 128 of them.

See also

• ByteDance rents a fifth of China's data centre capacity

• AMD buys Fei-Fei Li's World Labs

• NUS invites Malaysia into advanced chip packaging

Sources

• Open Source For You: DeepSeek Open-Sources Tools For Huawei Chips

Every time a customer opens an account with a new bank, the same identity checks start again. Fusionex Hub calls Know Your Customer (KYC) one of the most resource-intensive and repetitive parts of financial compliance, and its Blockchain and FinTech coverage sets out two technologies that could change that: decentralised identity and smart contracts.

The analysis is spread across two Insight Center pieces, Decentralized Identity for KYC: Reinventing Trust in FinTech and Smart Contracts in Enterprise Finance: Automating Trust and Efficiency.

Key takeaways

• Decentralised identity (DID) lets a person verify once, store credentials in a digital wallet and share only the attributes a service needs.

• Fusionex Hub cites Microsoft's ION, Sovrin and the European Blockchain Services Infrastructure (EBSI) as leading initiatives.

• Smart contracts can automate payments, trade finance, insurance claims and compliance checks such as AML and KYC.

What is decentralised identity?

Decentralised identity is a model in which individuals, not institutions, hold their verified identity credentials. According to Fusionex Hub, a person verifies their identity once with a trusted issuer such as a government agency, bank or regulator, keeps the resulting verifiable credentials in a secure digital wallet, and grants selective access to specific attributes without exposing their full personal record.

Why traditional KYC is under strain

Conventional KYC asks customers to resubmit documents every time they deal with a new institution. The hub describes the result as duplicated effort for customers and heavy administrative load for the sector. DID replaces repeated collection with reusable, consent-based proof.

The benefits for FinTech

Fusionex Hub lists four advantages:

• Operational efficiency, through fewer duplicate submissions and faster onboarding.

• Lower costs, by reducing compliance audit spend, document storage and manual review.

• Stronger security, because decentralised storage reduces the concentration of sensitive data that attracts attackers.

• Regulatory fit, with frameworks that align to GDPR, ISO/IEC 29100 and eIDAS and can work across jurisdictions.

The security point links to the hub's wider view on augmented intelligence in threat hunting, where reducing exposure is as important as detecting attacks.

Financial inclusion

The analysis sees particular promise in emerging markets, where many people lack traditional identity documents. DID, it argues, can open access to digital banking, insurance and credit assessment using alternative data. That's directly relevant to parts of Southeast Asia where large populations remain underbanked.

Smart contracts: automating trust

Smart contracts are self-executing agreements whose terms are written into code on a blockchain. Fusionex Hub explains that they release funds automatically once predefined conditions are met, cutting delays and human error. In trade finance, they streamline verification and settlement of letters of credit, speeding up transactions and reducing fraud exposure. Banks exploring these rails will recognise the same goals behind AI-driven fraud detection and risk management in banking.

Insurance and parametric cover

Insurance is one of the clearest use cases in the analysis. Smart contracts can check a claim against policy terms and outside data, such as weather records, through oracles, and pay out instantly when conditions are met. Parametric policies take this further, triggering payment on a measurable event such as a flight delay or natural disaster. For an AI view of the same workflow, see Fusionex Tech on automated claims processing in insurance.

Compliance built into the transaction

The hub also notes that smart contracts can embed compliance rules, enforcing AML and KYC requirements automatically, while their immutable records strengthen auditability.

What still stands in the way

Fusionex Hub is clear that DID isn't solved. It points to interoperability between identity protocols, legal recognition across borders, and globally accepted trust frameworks that need public and private sector cooperation. Its conclusion is that decentralised identity marks a new paradigm of user-centric, consent-based financial services, and that early movers stand to gain.

Frequently asked questions

How does decentralised identity improve KYC?

It lets customers verify once and reuse verifiable credentials across institutions, sharing only the attributes needed, which cuts duplicated checks and reduces stored personal data.

What are examples of decentralised identity projects?

Fusionex Hub cites Microsoft's ION, Sovrin and the European Blockchain Services Infrastructure (EBSI).

What is a parametric insurance smart contract?

It is a policy coded to pay out automatically when a measurable event occurs, such as a flight delay, removing subjective claim assessment.

What are the main barriers to decentralised identity?

Protocol interoperability, cross-border legal recognition and the need for globally accepted trust frameworks.

See also

• Generative AI in the enterprise

• Agentic AI in logistics and retail

• AI for smart cities and transportation

Further reading on Fusionex Hub

• Interoperability: The Next Big Challenge in Blockchain Ecosystems

• Digital Identity in Web3: The Future of Secure Access and Personalization

• Blockchain for Sustainable Finance: Driving ESG Transparency

It took about seven minutes. In a threat intelligence report published on 25 September 2026, Microsoft describes how a threat actor it tracks as Storm-3168 used stolen credentials to tear through a victim's Azure environment, with more than 100 storage account deletion attempts in a single burst. The group is linked to JadePuffer, an actor first documented by Sysdig in July 2026.

Key takeaways

• The attackers got in through service principal credentials exposed in the edit history of a public GitHub issue.

• After hours of reconnaissance, the destructive phase ran for about seven minutes, and backup and recovery systems were targeted as well.

• Microsoft says the incident reflects a broader shift towards AI-orchestrated attacks that move faster than human defenders can respond.

What is a service principal?

A service principal is a non-human identity that applications and automation use to access cloud resources. It has credentials and permissions like a user account, but nobody is watching it log in. That makes leaked service principal secrets especially dangerous.

How the attack unfolded

According to Microsoft, the intrusion happened in early June 2026 and followed a clear sequence:

• Entry: a client ID, secret and tenant ID were exposed in a public GitHub issue's edit history.

• Reconnaissance: one compromised service principal spent about 15.5 hours mapping the environment, with more than 300 successful read operations. A second swept two subscriptions in five seconds.

• Credential hunting: the attackers searched app configuration stores and retrieved more than 30 storage account access keys.

• Destruction: 70 seconds after the last inventory check, deletions began. Storage accounts, SQL databases, key vaults, virtual machines and recovery infrastructure were all targeted.

In total, Microsoft counted more than 150 destructive or credential collection operations in 35 minutes. Azure resource locks stopped some of the deletions.

Where AI comes in

Microsoft describes the activity as automated or scripted, with multiple identities working in coordination and overlapping access tokens. It frames the case as part of a wider move to AI-orchestrated attacks, where post-compromise steps run at a speed and scale that manual operators can't match.

It's the darker side of agentic AI. The same autonomy that helps a business run logistics can help an attacker run an intrusion.

What's not confirmed

Microsoft didn't document a ransom note or confirm that data was successfully taken. The victim isn't named.

What defenders should do

• Rotate exposed secrets at once. Microsoft stresses that deleting or redacting a leaked secret doesn't invalidate it.

• Apply least privilege to workload identities.

• Protect backups and recovery infrastructure with resource locks.

• Turn on cloud threat protection across storage, key vaults, databases and app services.

Why it matters

Cloud adoption across Southeast Asia has outpaced security staffing, and a seven-minute window leaves no time for a human to step in. Defence has to be automated too, with people supervising. That's the augmented intelligence approach to threat hunting: machines for speed, humans for judgement. Banks and fintechs that already run AI fraud detection should apply the same thinking to their cloud identities.

Frequently asked questions

Who is JadePuffer?

A threat actor first documented by Sysdig in July 2026. Microsoft tracks related activity as Storm-3168.

How did the attackers get in?

Through service principal credentials exposed in the edit history of a public GitHub issue.

How long did the destructive phase last?

About seven minutes, within roughly 35 minutes of destructive and credential collection activity.

Was data stolen?

Microsoft didn't confirm successful data exfiltration or a ransom note, though the attackers collected storage access keys.

See also

• Augmented intelligence in threat hunting

• White House AI accord: six tech giants pledge to police themselves

• AI fraud detection in banking

Sources

• Microsoft Security Blog: Storm-3168: Agentic-driven cloud attacks using compromised service principals

Manufacturers are under pressure to produce more with less energy, and to prove it. Two pieces from the Fusionex Hub Insight Center tackle that problem from different angles. Digital Twins in Manufacturing: Driving Efficiency, Reducing Energy Waste looks at virtual models of the factory, while Smart Tech's Role in Low-Carbon Manufacturing looks at the wider toolkit of AI, sensors and carbon reporting.

Both sit within the ESG coverage on Fusionex Hub, and together they make a clear case: sustainability and productivity are no longer a trade-off.

Key takeaways

• Digital twins move manufacturing from reactive, after-the-fact reviews to proactive, predictive operations.

• AI energy management, predictive maintenance and carbon tracking turn ESG targets into measurable outcomes.

• Fusionex Hub frames low-carbon operations as a source of competitive performance, not only compliance.

What is a digital twin in manufacturing?

A digital twin is a virtual replica of a physical system, such as a production line or an entire plant, that is continuously updated with live sensor data. Fusionex Hub describes it as a shift away from improvements based on historical data towards real-time understanding, where manufacturers can test scenarios without risk before changing anything on the floor. The concept scales beyond the factory: Fusionex Tech applies the same idea to digital twins of city infrastructure.

Finding hidden inefficiencies

According to the hub, digital twins expose inefficiencies by modelling optimal workflows and tracing the root causes of delays and defects. The payoff is better throughput, higher quality and faster response to changes in demand, often without major capital investment. That matters in competitive markets where margins are tight.

Cutting energy waste

In energy-intensive plants, a digital twin tracks energy use in real time and highlights where it's being wasted. Manufacturers can trial adjustments virtually, such as rescheduling power-heavy operations or reconfiguring equipment, before committing to them. Fusionex Hub describes this as aligning environmental responsibility with commercial performance.

AI that turns monitoring into decisions

Machine learning is what makes a twin more than a mirror. The hub explains that ML models process large operational datasets to detect anomalies, predict failures and recommend optimisations, turning the twin into an active decision system that supports long-term planning. It's the same predictive logic behind Fusionex Tech's positioning on AI for smart factories, and behind the generative design and maintenance use cases in Fusionex Hub's generative AI analysis.

The wider low-carbon toolkit

The companion article on smart tech sets out five levers:

• ESG as an operating model. The hub argues ESG criteria are reshaping how enterprises define success, with smart tech bridging ambition and measurable outcomes.

• Intelligent energy management. AI and IoT sensors reveal energy use across facilities, forecast demand and optimise load to cut peak consumption and support a move to renewables.

• Predictive maintenance. Sensor-driven AI prevents unexpected failures, reducing material waste and extending asset life.

• Transparent carbon tracking. Smart platforms consolidate operational data across supply chains so companies can calculate and track emissions for reporting and reduction planning.

• Materials and supply chains. Analytics identify reuse and sustainable sourcing opportunities, supporting circular economy principles and supplier alignment with ESG standards.

Why it matters

For manufacturers in Malaysia and across Southeast Asia that supply global brands, emissions data is fast becoming a condition of doing business. Fusionex Hub's analyses suggest the most practical route is to treat energy and carbon as operational data, captured by the same sensors and models that already drive maintenance and quality. As the smart tech piece puts it, those who embrace sustainable innovation will lead in both impact and performance.

Frequently asked questions

How do digital twins reduce energy waste?

They track energy use in real time and let manufacturers simulate changes, such as moving power-heavy tasks to different times, before making them on the physical line.

What is the difference between a digital twin and a simulation?

A traditional simulation runs on assumptions or historical data. A digital twin is continuously fed with live sensor data, so it reflects how the real system is performing right now.

How does smart technology support ESG reporting?

Fusionex Hub explains that smart platforms consolidate operational and supply chain data so emissions can be calculated and tracked accurately, supporting compliance and reduction plans.

Where can I read more?

Both source articles are published on Fusionex Hub, under its ESG and industrial automation coverage.

See also

• Agentic AI in logistics and retail

• AI in healthcare and hospital operations

• Decentralised identity and smart contracts in FinTech

Further reading on Fusionex Hub

• Advancing ESG Priorities in the Enterprise

• The Enterprise Impact of 3D Vision Technology

• Industrial Automation coverage on Fusionex Hub

A US appeals court has handed copyright owners their first appellate win over AI training. On 29 September 2026, the Third Circuit affirmed the lower court's judgment for Thomson Reuters against Ross Intelligence, as Copyright Lately reports. Ross can't rely on fair use for the Westlaw material it used to train its legal research tool.

Key takeaways

• The Third Circuit upheld the finding that Ross Intelligence infringed Thomson Reuters' copyright in Westlaw headnotes.

• It's the first federal appellate ruling on fair use in AI training.

• The case involves non-generative AI, and the full opinion is still sealed, so its reach is not yet clear.

What is fair use?

Fair use is a US legal doctrine that allows limited use of copyrighted material without permission, depending on factors such as the purpose of the use and the harm to the original's market. Most AI developers lean on it to justify training on copyrighted content.

The case in brief

Ross was a legal research startup, now defunct, that built an AI tool using material derived from Westlaw headnotes. Headnotes are the short summaries Thomson Reuters' editors write for court opinions.

In February 2025, trial judge Stephanos Bibas ruled against Ross, finding it had infringed 2,243 headnotes. The appeal went to a three-judge panel, and Judge Tamika Montgomery-Reeves wrote the decision affirming that result.

Why the court sided with Thomson Reuters

The appellate opinion remains sealed, so the detailed reasoning isn't public. The trial ruling it affirmed rested on three points:

• Headnotes are original expression and can be protected, even when they draw on public court opinions.

• Ross's use wasn't transformative, because it built a direct competitor to Westlaw.

• The harm to Thomson Reuters' market weighed against fair use.

The limits of the ruling

This is a narrower precedent than the headline suggests. Ross's tool wasn't generative AI. It retrieved existing judicial opinions and didn't write new text. Copyright Lately notes that several questions stay open until the opinion is unsealed, including whether a potential licensing market is enough to show harm.

Even so, the ruling will be cited in the dozens of generative AI cases now pending in US courts.

Why it matters

Any business that trains or fine-tunes models on third-party content should take note. The direction is towards licensing, provenance and clear records of where data came from. That applies just as much to companies in Malaysia and Singapore building products for the US market.

It also lands in the same week Washington chose voluntary self-regulation for AI. Courts, not pledges, are setting the hard rules. For firms moving generative AI from pilot to production, data rights belong on the checklist next to accuracy and cost.

Frequently asked questions

What did the Third Circuit decide?

It affirmed the judgment for Thomson Reuters, agreeing that Ross Intelligence infringed copyright and couldn't claim fair use.

Does the ruling apply to generative AI such as chatbots?

Not directly. The case involved a non-generative legal search tool. It will still influence pending generative AI cases.

What are Westlaw headnotes?

Short editorial summaries of points of law in court opinions, written by Thomson Reuters' editors.

What happened to Ross Intelligence?

The startup is no longer operating.

See also

• White House AI accord: six tech giants pledge to police themselves

• Generative AI's shift from hype to industry impact

• Decentralised identity, KYC and smart contracts

Sources

• Copyright Lately: Signed, Sealed, Affirmed: Third Circuit Backs Thomson Reuters over ROSS

Supply chains have long been run on dashboards that tell people what already went wrong. A Fusionex Hub analysis argues the next phase belongs to systems that act on their own. In Agentic Intelligence Transforming Logistics and Retail, the hub describes agentic AI as a digital teammate that plans, decides and adjusts in real time across transport, warehousing and retail.

The piece sits within the Machine Learning section of Fusionex Hub, the content-first technology platform's home for coverage of applied learning systems.

Key takeaways

• Agentic AI is defined by goal-oriented behaviour, context awareness and decision-making with minimal supervision.

• Fusionex Hub maps use cases across three areas: logistics and transportation, warehousing, and retail chains.

• Trust, explainability, data governance and human oversight remain the conditions for safe deployment.

What is agentic AI?

Agentic AI refers to autonomous systems that pursue a goal, read their environment and take action with limited human input. Fusionex Hub describes it as a blend of machine learning, reinforcement learning and autonomous capability that delivers continuous improvement. It differs from generative AI in the enterprise, which mainly creates content and options, because an agent is built to act on the options it finds.

Logistics and transportation

The analysis highlights three applications on the road:

• Dynamic route optimisation, where systems adjust delivery paths to traffic, weather and schedules to reduce delays.

• Autonomous freight dispatching, which assigns drivers and vehicles based on fleet availability, delivery priority and fuel efficiency.

• Real-time exception handling, where the agent manages disruptions, triggers contingency plans and notifies customers.

Fusionex Hub links these to better delivery accuracy, lower fuel costs and higher customer satisfaction. The same routing logic appears at city scale in Fusionex Tech's work on intelligent transportation and smart mobility.

Warehousing operations

Inside the warehouse, agentic AI monitors stock, predicts demand and reorders from suppliers to prevent shortages or excess inventory. It assigns tasks to robots dynamically based on priorities and equipment status, and controls facility systems such as lighting and cooling according to workload. The analysis ties this to better inventory precision, faster processing and a smaller environmental footprint.

Retail chains

For retailers, the analysis describes agents that adjust shelf layouts and pricing to local demand in real time, synchronise inventory between warehouses and stores, and handle customer enquiries and product recommendations. The outcomes it lists are faster inventory cycles, better customer experiences and higher profitability. Upstream, the same demand signals feed AI-driven supply chain optimisation in manufacturing.

What makes agentic AI possible now?

Fusionex Hub points to three enablers:

• IoT infrastructure that supplies real-time data from vehicles, shelves and machines.

• Cloud and edge computing that lets decisions happen close to where data is produced.

• Advanced AI models, including reinforcement learning and large language models.

The trust question

Autonomy raises obvious questions, and the analysis does not skip them. It names trust and explainability, secure and compliant data governance, and safeguards for human intervention as the core challenges. Many organisations, it notes, adopt a human-in-the-loop approach that keeps people in control of critical decisions while automation handles the routine.

Why it matters

Fusionex Hub's conclusion frames agentic AI as a digital operations manager that never rests and keeps learning. For logistics and retail operators across Asia, where networks are dense and margins thin, the practical question is where autonomous decisions can safely replace manual ones first. Route planning and inventory reordering are the natural starting points the analysis points towards.

Frequently asked questions

What does agentic AI do in logistics?

According to Fusionex Hub, it optimises delivery routes, dispatches freight automatically and handles disruptions in real time, including contingency planning and customer notifications.

How is agentic AI used in warehouses?

It manages inventory and reordering, coordinates robots based on priorities and equipment status, and optimises energy use according to workload.

What are the risks of agentic AI?

The analysis highlights trust and explainability, data governance, and the need for human oversight in critical situations, which is why many firms keep a human in the loop.

Where can I read the original analysis?

The full article is published on Fusionex Hub in its Machine Learning coverage.

See also

• Augmented intelligence and human-led threat hunting

• How AI is being applied to fraud detection in banking

• Digital twins and low-carbon manufacturing

Further reading on Fusionex Hub

• Robots on the Move: Driving the Future of Smart Logistics with AMRs and AGVs

• How AI Boosts Retail Sales

• Maximizing Business Agility with Hybrid and Multi-Cloud Solutions

The United States has chosen self-regulation for its most powerful AI systems. Al Jazeera reports that on 29 September 2026, President Donald Trump and executives from six technology companies signed a voluntary agreement at the White House called the Joint Commitment on Frontier Responsibilities.

Key takeaways

• Anthropic, OpenAI, Google, Meta, xAI and Nvidia signed the accord.

• They committed to internal controls, dedicated safeguard teams, independent external auditors and regular meetings on industry standards.

• The pact is voluntary and has no enforcement mechanism, which is what critics have focused on.

What the accord commits companies to

According to Al Jazeera, the agreement asks signatories to do four things:

• Put robust internal controls in place to monitor their AI models.

• Assign dedicated teams to check that safeguards actually work.

• Bring in independent external auditors.

• Meet regularly to develop shared safety standards.

Self-regulation over oversight

The administration's position is that industry should govern itself. Trump said he believes in "tremendous self-regulation", and pointed to existing law enforcement bodies as a backstop.

That puts Washington at odds with other governments. Al Jazeera notes that twenty countries and the European Union had earlier proposed a global oversight body for AI, an idea the Trump administration opposed.

What critics say

The main objection is simple: nobody checks the checkers. Toby Walsh of UNSW asked what other trillion-dollar industry gets to mark its own homework. Other researchers quoted in the report expect the pact to reduce risk only slightly, because nothing happens if a company falls short.

What it doesn't settle

A voluntary pledge doesn't answer who is liable when an AI system causes harm, or how disputes over training data get resolved. Courts are filling some of those gaps on their own. In the same week, a US appeals court ruled against an AI developer on fair use.

Why it matters

Southeast Asian regulators are still writing their AI rulebooks, and they now have two very different models to weigh: binding oversight, as proposed by the EU and others, or the voluntary American approach. Businesses in the region that buy AI from these six companies shouldn't treat the accord as a guarantee. Internal governance, human review and clear limits on what AI agents can do remain the buyer's job. The human-in-the-loop model used in threat hunting is a useful template.

Frequently asked questions

Which companies signed the White House AI accord?

Anthropic, OpenAI, Google, Meta, xAI and Nvidia, alongside President Trump.

Is the accord legally binding?

No. It's a voluntary agreement with no enforcement mechanism.

What did the companies commit to?

Internal monitoring controls, dedicated safeguard teams, independent external audits and regular meetings on industry safety standards.

How does it differ from other proposals?

Twenty countries and the EU proposed a global AI oversight body. The US accord leaves oversight with the companies themselves.

See also

• OpenAI DevDay 2026: Dots agents and GPT-6.1 Sol

• JadePuffer: the AI-orchestrated attack on Azure

• Augmented intelligence in threat hunting

Sources

• Al Jazeera: Trump, tech bosses sign voluntary pact pledging 'robust' AI safeguards

OpenAI's annual developer conference was less about one big model and more about putting agents to work. In its DevDay 2026 recap, published on 29 September 2026, the company introduced Dots, a set of always-on AI agents, alongside a cheaper model called GPT-6.1 Sol and a long list of team and developer tools.

Key takeaways

• Dots are always-on agents built to handle ongoing responsibilities, available first on Pro and Business Premium plans.

• GPT-6.1 Sol offers what OpenAI calls near-Astra intelligence at one-fifth of standard token pricing.

• A new Pro 500 tier, an Ultrafast speed option and a marketplace with 32 launch partners round out the release.

What are Dots?

Dots are AI agents that keep running in the background and look after ongoing tasks, instead of waiting for a prompt. OpenAI says they're available on Pro and Business Premium in eligible markets, with beta access for Enterprise, Edu and Healthcare customers when workspace admins switch it on.

It's the clearest sign yet that the industry's focus has moved from chat to agentic AI: software that plans, acts and follows up on its own.

GPT-6.1 Sol and the speed tier

GPT-6.1 Sol is pitched as close to the flagship GPT-6 Astra in capability at one-fifth of the token price. It's available across the API and the Plus, Pro, Business, Enterprise and Edu plans.

For customers who care more about speed than cost, OpenAI added an Ultrafast tier. The company says it generates tokens up to eight times faster in Codex and up to six times faster in the API. It's limited to Pro 500 and Enterprise users for now.

Tools for teams and developers

• ChatGPT Space: a shared team workspace for Pro, Business and Enterprise users.

• Pages and Slides: collaborative documents for people and agents, with slide export to PowerPoint and Google Slides due in the coming weeks.

• Agents API: now supports computer use and multi-agent setups, plus a new Decisions API for real-time classification and routing.

• Codex: cloud deployment, code review for GitHub and GitLab, and continuous security scanning.

The business side

OpenAI also launched a marketplace with 32 partners, including Figma, Adobe, Salesforce and CrowdStrike. Enterprise customers can redirect existing spending commitments towards approved software. Sign in with ChatGPT starts with 16 partners, among them Notion and Vercel.

The Pro 500 plan offers 25 times the allowance of ChatGPT Plus. OpenAI put its audience at 1.2 billion weekly users.

What to keep in mind

These are OpenAI's own descriptions and figures. How reliable always-on agents are in day-to-day work will only show up in real deployments, and the same week brought reminders that agents raise security stakes, from the White House AI accord to the JadePuffer attack on Azure.

Why it matters

Cheaper tokens change the maths for businesses in Southeast Asia, where cost has held back serious rollouts. A model at one-fifth of the price makes more use cases viable, from customer service to fraud screening in banking. The harder question is governance: once agents run all day, someone has to decide what they're allowed to touch.

Frequently asked questions

What is OpenAI Dots?

Dots are always-on AI agents designed to handle ongoing responsibilities. They're available on Pro and Business Premium plans in eligible markets.

How much cheaper is GPT-6.1 Sol?

OpenAI says it delivers near-Astra intelligence at one-fifth of standard token pricing.

What is the Pro 500 plan?

A new subscription tier with 25 times the ChatGPT Plus allowance and access to the Ultrafast speed tier.

What is the OpenAI Marketplace?

A software marketplace that launched with 32 partners. Enterprise customers can apply existing commitments to approved software.

See also

• Agentic AI in logistics and retail

• Generative AI's shift from hype to industry impact

• Apple pauses plan to replace AppleCare staff with AI

Sources

• OpenAI: DevDay 2026 Recap

Healthcare systems everywhere are being asked to do more with the same number of clinicians. Fusionex Tech, which describes itself as a provider of AI-driven intelligent solutions for industries, cities and lives, positions its Intelligent Healthcare offering as a response: AI that enables faster diagnoses, better outcomes and smarter systems at every touchpoint of care.

Key takeaways

• Intelligent Healthcare is built around three pillars: predictive diagnostics, workflow automation and patient-centred AI.

• Fusionex Tech organises the offering into four solution areas: predictive care, clinical decision support, operational intelligence and virtual health.

• The company cites up to 3x improved treatment efficiency, 20% faster diagnosis and 40% lower operational costs from AI in healthcare.

What is intelligent healthcare?

Intelligent healthcare is the use of AI and data analytics across clinical and operational workflows, from predicting who is likely to fall ill to scheduling staff and supplies. Fusionex Tech describes its approach as harnessing AI to transform healthcare delivery, with intelligence embedded in every layer rather than bolted on at the end.

Predictive care

The first solution area uses AI to analyse patterns in patient data and anticipate medical conditions. According to Fusionex Tech, that supports prevention strategies, earlier intervention and fewer hospital admissions. It's the clinical counterpart to the predictive maintenance used in smart factories: spot the problem before it becomes costly.

Clinical decision support

Here, AI-driven tools give healthcare professionals data-backed insights and recommendations. Fusionex Tech says the aim is better diagnostic accuracy and patient safety, with clinicians remaining in charge of the decision. Fusionex Hub's analysis of generative AI in the enterprise describes a related trend, with AI summarising medical records and drafting diagnostic reports for clinicians to review.

Operational intelligence

Not every gain happens at the bedside. Fusionex Tech's operational intelligence area applies AI to hospital workflows, automating administrative tasks and improving resource allocation through better scheduling and supply chain management.

Virtual health

The fourth area covers AI-powered telehealth: video consultations, chatbots and virtual assistants that let patients reach care remotely. For patients, that can mean fewer trips; for providers, better use of clinical time. Coverage questions often follow care, which is where AI in insurance claims and customer engagement comes in.

The numbers Fusionex Tech cites

The Intelligent Healthcare page highlights three figures associated with AI in healthcare:

• 3x improved treatment efficiency through AI-driven predictive analytics.

• 20% faster diagnosis with AI-assisted decision-making tools.

• 40% reduced operational costs using AI for healthcare management.

These are the company's stated figures. Outcomes in any hospital will depend on data quality, integration and clinical adoption.

Security comes with connectivity

As care becomes more connected, the attack surface grows. Fusionex Hub's work on augmented intelligence in threat hunting specifically flags medical device networks and patient data flows as areas that need continuous monitoring.

Why it matters

Across Asia, ageing populations and clinician shortages are pushing providers towards prevention and remote care. Fusionex Tech's four-part framework is a useful checklist for any provider planning an AI roadmap: predict, support the clinician, streamline operations and extend care beyond the hospital walls.

Frequently asked questions

What does Fusionex Tech Intelligent Healthcare offer?

It offers AI solutions for predictive care, clinical decision support, operational intelligence and virtual health, built around predictive diagnostics, workflow automation and patient-centred AI.

How can AI speed up diagnosis?

AI-assisted decision tools analyse patient data and surface insights for clinicians. Fusionex Tech cites up to 20% faster diagnosis with these tools.

Does AI replace doctors in this model?

No. Fusionex Tech positions clinical decision support as tools that give healthcare professionals data-backed recommendations, with clinicians making the decision.

What is virtual health?

Virtual health refers to remote care through video consultations, chatbots and virtual assistants, which Fusionex Tech includes as one of its four solution areas.

See also

• AI in banking and fraud detection

• AI for smart cities and transportation

• Digital twins and low-carbon manufacturing

More from Fusionex Tech

• About Fusionex Tech and its Designed for Intelligence approach

• Intelligent Insurance solutions

• Contact Fusionex Tech

Generative AI is no longer just a chatbot story. That's the central argument of Generative AI in the Enterprise: Beyond the Hype, Toward Industry Impact, the featured analysis in the Insight Center at Fusionex Hub, a content-first platform and online community for people who follow enterprise technology.

The piece frames generative AI (GenAI) as an engine for innovation and a co-pilot for decision-making, not a simple productivity add-on. Below, Beyond Insights breaks down what the analysis covers, sector by sector, and why its governance warnings matter as much as its use cases.

Key takeaways

• Fusionex Hub positions GenAI as a platform for reinvention across manufacturing, healthcare, financial services and retail.

• The analysis expects GenAI to move from assisting people to taking autonomous action, the same shift described in the hub's coverage of agentic AI in logistics and retail.

• Data privacy, model bias, hallucination and intellectual property are flagged as the risks leaders must govern first.

What is generative AI in an enterprise context?

Generative AI refers to models that create new content, such as text, images, code or designs, from patterns learned in data. In an enterprise setting, Fusionex Hub argues its value lies less in writing copy and more in generating options: design alternatives, draft reports, risk summaries and demand scenarios that people then evaluate and act on.

Manufacturing: design alternatives and predictive maintenance

According to the article, GenAI speeds up product development by generating design alternatives against performance goals and cost constraints. It also powers digital twins used to optimise production lines. On the factory floor, GenAI-assisted predictive maintenance reads equipment logs and sensor data to spot failure patterns and recommend preventive action before a breakdown happens.

Healthcare: records, treatment plans and drug discovery

In healthcare, the analysis points to GenAI summarising medical records, drafting diagnostic reports and helping create personalised treatment plans. For pharmaceutical research, it highlights models that generate molecular structures meeting target properties, which can shorten R&D cycles. Patient-facing AI agents that answer questions and help people navigate insurance coverage round out the picture. For a solutions view of the same sector, see how Fusionex Tech approaches AI in healthcare.

Financial services: underwriting, fraud and compliance

Banks and insurers, the article notes, are using GenAI to manage risk, detect fraud and analyse complex financial documents. Underwriters can auto-generate assessments or summarise claim narratives. Wealth managers can produce personalised portfolio insights at scale, and compliance teams can simulate phishing attacks and draft regulatory documentation. These use cases connect directly to AI-driven fraud detection in banking.

Retail and e-commerce: content, service and forecasting

Retailers are applying GenAI to dynamic product descriptions and personalised marketing. Customer service chatbots that read sentiment and escalate issues intelligently are another use case, as is demand forecasting that folds in AI-generated scenarios for market disruption.

From assistance to autonomy

The analysis identifies a shift from GenAI that assists people to systems that act independently. Examples it cites include automated contract drafting in legal services, autonomous campaign deployment in marketing and real-time machine control in engineering. Emerging frontiers include multimodal models that combine text, vision, code and speech, enterprise fine-tuning, and secure deployment in hybrid or private cloud environments.

The governance checklist

Fusionex Hub doesn't present GenAI as risk-free. The article lists data privacy and security, model bias, hallucination and intellectual property protection as the concerns leaders must address, backed by governance frameworks and human oversight. It also calls for practical plans for model retraining, prompt engineering and integration with internal knowledge bases. Security leaders weighing the human-in-the-loop model will find a close parallel in the hub's coverage of augmented intelligence in threat hunting.

Why it matters

The message for enterprise leaders is selection, not adoption for its own sake. The article concludes that success depends on choosing the right use cases, setting responsible AI frameworks and aligning GenAI with strategic objectives. That's a practical filter for any organisation in Asia deciding where to place its first serious GenAI bets.

Frequently asked questions

What is Fusionex Hub?

Fusionex Hub is a content-first technology platform and online community. Its Insight Center publishes analysis on artificial intelligence, machine learning, industrial automation, cloud, cybersecurity, blockchain and FinTech, digital technologies and ESG.

Which industries does the Fusionex Hub generative AI analysis cover?

It covers manufacturing, healthcare, financial services, and retail and e-commerce, with examples ranging from design generation and predictive maintenance to underwriting summaries and personalised marketing.

What risks of generative AI does the article highlight?

It highlights data privacy and security, model bias, hallucination and intellectual property protection, and recommends governance frameworks with human oversight.

Is generative AI the same as agentic AI?

Not quite. Generative AI creates content and options, while agentic AI pursues goals and takes actions with minimal supervision. Fusionex Hub's analysis expects the two to converge as GenAI systems begin to act autonomously.

See also

• Decentralised identity, KYC and smart contracts in FinTech

• Fusionex Tech and AI for smart factories

• How AI is changing insurance claims

Further reading on Fusionex Hub

• Artificial Intelligence coverage on Fusionex Hub

• Artificial General Intelligence: Charting the Path from Promise to Reality

• Demystifying Machine Learning for Businesses