OpenAI Slows Advanced AI Development After Cyberattack

The OpenAI logo in this illustration taken June 11, 2026. (Reuters)
The OpenAI logo in this illustration taken June 11, 2026. (Reuters)
TT

OpenAI Slows Advanced AI Development After Cyberattack

The OpenAI logo in this illustration taken June 11, 2026. (Reuters)
The OpenAI logo in this illustration taken June 11, 2026. (Reuters)

ChatGPT creator OpenAI said Tuesday that it was tapping the brakes on development of its most advanced AI model and tightening internal controls, a month after revealing a cyberattack carried out by one of its AI tools.

OpenAI is a key player in the rapid global buildout of artificial intelligence infrastructure and tools that some have likened to an arms race.

The company said in a blog post on Tuesday that it was holding off on conducting the biggest AI training run it had ever planned while it checks that the resulting model would behave as expected.

Training runs are computationally intense exercises where models are fed enormous amounts of text and images.

This combined with fine-tuning billions of internal settings results in their abilities to reason and respond to prompts and other inputs.

"We always said we would take action if we felt that model capabilities were outstripping the pace of safety and alignment," OpenAI CEO Sam Altman said.

In mid-July, an AI agent based on two OpenAI models left its confined testing environment on its own initiative to venture onto the internet and attack Hugging Face, a platform where developers around the world share their AI models.

Similarly, OpenAI rival Anthropic revealed in late July that three of its models undergoing testing had also carried out unauthorized intrusions into the computer systems of three organizations.

The incidents prompted a petition signed by more than 1,000 tech industry employees calling on the US government to support a coordinated slowdown in the development of the most advanced AI systems.

OpenAI had halted training of its latest models for two weeks before resuming it under tighter controls.

Much of the work related to Astra -- the company's next major model -- remains suspended, as the company determined in early August that the model could cross the warning threshold it has set for itself regarding the hacking capabilities of its AI systems.

Given that risk, OpenAI's internal rules require it to create stronger safeguards before development can resume.

OpenAI also said Tuesday that it was developing a new system to peer into the internal reasoning of models and sound the alarm to humans within 30 minutes of suspicious behavior.

That monitoring, however, will require an additional 20 percent more in computing power.

OpenAI's own research in 2025 showed the limits of this approach: a model that knows it is being monitored can learn to conceal its intentions in its reasoning.

The company has been promising a detailed technical account of the Hugging Face incident, but has yet to publish it.

Tuesday's blog post said it would be released "in the coming weeks."



China Robot Makers Flock to Beijing Show, Seek Path to Mass Adoption

 Two robots produced by Unitree take part in a fighting demonstration during the 2026 World Robot Conference in Beijing on August 19, 2026. (AFP)
Two robots produced by Unitree take part in a fighting demonstration during the 2026 World Robot Conference in Beijing on August 19, 2026. (AFP)
TT

China Robot Makers Flock to Beijing Show, Seek Path to Mass Adoption

 Two robots produced by Unitree take part in a fighting demonstration during the 2026 World Robot Conference in Beijing on August 19, 2026. (AFP)
Two robots produced by Unitree take part in a fighting demonstration during the 2026 World Robot Conference in Beijing on August 19, 2026. (AFP)

China's robot makers showcased their latest machines at a Beijing conference on Wednesday, seeking to convince investors and customers that the technology is moving forward for broader commercial adoption.

More than 300 companies are attending the World Robot Conference which runs through Sunday, displaying over 2,000 exhibits and launching more than 150 products, according to Beijing authorities.

The conference comes as robotics emerges as a new front in the US-China rivalry and as investor enthusiasm for the sector hits fresh highs in China.

Shares in Unitree, China's best-known humanoid robot ‌maker, soared ‌nearly six-fold in their Shanghai trading debut on Wednesday.

ROBOTICS AN 'IMPORTANT FORCE' ‌IN ⁠CHINA'S DEVELOPMENT

Following its ⁠successful IPO, which was more than 8,000 times oversubscribed by retail investors, startup Lumos Robotics and the robotics division of China's largest auto exporter Chery Automobile told Reuters they were also considering stock market listings.

At the conference's opening ceremony, Xin Guobin, vice minister of the Ministry of Industry and Information Technology, pledged support for the sector's development, saying robotics had already become "an important force" in China's economic and social development, state-backed financial news outlet Cailianshe reported.

While robots in ⁠China are increasingly being deployed for tasks, such as hotel food ‌deliveries and on some assembly lines, large-scale adoption beyond ‌pilot projects remains limited.

Startup Robotera displayed a slick demonstration of a parcel-sorting humanoid robot torso ‌on a wheeled tripod base at the show. The robots have a continuous running ‌time of 900 to 1,000 hours, and are 85% as effective as a human worker, a sales representative told Reuters.

COMPETITION HELPS INDUSTRY'S GROWTH

The industry's rapid expansion mirrors China's electric vehicle sector a few years ago, when scores of companies emerged rapidly and competed aggressively.

"A large number of (robot) companies appeared ‌almost overnight," Zhang Guibing, head of AiMOGA Robotics, Chery's robotics division, told Reuters on the sidelines of the conference.

"But from the perspective ⁠of industry development, ⁠there is a positive side to this. It allows everyone to compete fully," he said. "In the end, this will help deepen the industry's technology and capabilities. From that perspective, I think this is a good thing."

Global humanoid robot shipments nearly quadrupled in the first half of this year to 19,100 units, with Chinese manufacturers dominating the market, according to data from Smart Analytics Global.

Still, Georg Stieler, a robotics analyst who advises industrial companies in China, estimates that between 50% and 70% of humanoid robots produced this year could end up in "data factories", where they are used to collect training data rather than perform productive work for paying customers.

"Competition in the Chinese market could become extremely intense in the future, especially when it comes to costs," Zhang said, adding humanoid robots could take about a decade to mature before seeing rapid adoption.


Microsoft Arabia Chief Outlines Saudi Arabia's AI Priorities for Next Phase

Ayman Al-Ghamdi, President of Microsoft Arabia (Company)
Ayman Al-Ghamdi, President of Microsoft Arabia (Company)
TT

Microsoft Arabia Chief Outlines Saudi Arabia's AI Priorities for Next Phase

Ayman Al-Ghamdi, President of Microsoft Arabia (Company)
Ayman Al-Ghamdi, President of Microsoft Arabia (Company)

After years of building infrastructure and regulatory frameworks, Saudi Arabia's challenge is shifting toward how to effectively integrate artificial intelligence into the core operations of institutions. This is the gap Ayman Al-Ghamdi, the new president of Microsoft Arabia, has placed at the top of his priorities as he takes over the company's business in the Kingdom.

In his first media interview since taking office last July, Al-Ghamdi identified three priorities for the next phase: building a trusted cloud foundation, moving AI from experimentation to practical deployment, and expanding local capabilities.

This comes alongside the launch of the Azure region in Saudi Arabia's Eastern Province, the expansion of skills programs, and the development of scalable local solutions that can grow regionally and globally.

The implementation gap within institutions

Al-Ghamdi says Saudi Arabia has completed much of the foundational work required to move toward an economy that relies more heavily on data and AI, pointing to progress in infrastructure, data governance, responsible AI, and capacity building.

In this context, he cites what he described as the progress made by the Saudi Data and Artificial Intelligence Authority, noting that the regulatory ecosystem includes 32 regulatory instruments related to data and 13 related to AI. However, he believes the next phase will depend more heavily on what happens within institutions themselves.

He tells Asharq Al-Awsat: "Many institutions are still treating AI as something to experiment with, rather than integrating it into core operations."

In his view, closing this gap requires linking AI directly to business and service performance and customer experience, while providing the appropriate data, governance, talent, and operating models capable of turning adoption into measurable impact.

This approach forms the central focus of Al-Ghamdi's priorities for his first year. He believes the goal is to make the impact of AI "real, not just a headline," so that it is reflected in how institutions operate, the services they provide, and the results they achieve.

Three priorities for the first year

Al-Ghamdi's first priority is what he calls a "trusted cloud foundation," particularly as the Azure region in the Kingdom's Eastern Province prepares to become available for customer workloads. He links this step to institutions' ability to modernize critical systems securely and responsibly, rather than simply adding more computing capacity inside the Kingdom.

The second priority is moving AI projects from experimentation to practical value. He says the measure of success will not be the number of use cases companies announce, but the extent of improvement achieved in productivity, services, operations, and customer experience.

He points to the O3ai platform, an intelligent system for smart factories developed by Obeikan Group, as an example of what this transition from experimentation to practical use could look like, according to his response.

The third priority is building Saudi capabilities, based on Microsoft's commitment to help 3 million people in Saudi Arabia acquire AI skills by 2030. He says the objective is for the transformation to be built and sustained locally, because "technology alone does not create transformation, people do."

The cloud region... more than local hosting

Al-Ghamdi does not present the launch of the Saudi cloud region simply as a matter of capacity or data residency. Instead, he links it to the ability of companies and government entities to modernize critical workloads and prepare to expand their use of AI.

He notes that expansion at the national level requires an integrated set of elements, including cloud readiness, data infrastructure, security, governance, skills, and a strong partner network. From this perspective, he believes the broader impact of the new region could emerge through a greater share of digital value creation shifting into the Kingdom.

He says this could mean more local solutions, deeper technical capabilities, greater innovation among partners, and, over the longer term, the production of intellectual property capable of serving the Saudi market and other markets in the region.

A more competitive market

Al-Ghamdi views the growing range of choices available to Saudi customers among local and global cloud service providers as a positive development, considering it a reflection of a "strong and rapidly maturing" market. As cloud capacity expands and access to advanced AI models becomes easier, he does not believe competition will remain focused solely on who owns the infrastructure or the model.

He notes that "differentiation will shift from access to impact," explaining that the question will not simply be who has the model, but who can help institutions use it safely and responsibly in ways that genuinely change how they work.

He links this to an effort to bring cloud, data, cybersecurity, productivity tools, business applications, developer tools, and model options together on a single platform, making AI part of daily workflows, decisions, operations, and customer experiences.

He also emphasizes control over the data, context, and institutional knowledge that distinguish each organization. Al-Ghamdi sums up the elements of competition on which Microsoft is betting in three concepts: "trust, choice, and control," alongside the ability to turn AI into measurable impact at scale.

Uneven returns from AI

Despite the acceleration of investment in generative AI, Al-Ghamdi acknowledges the risk that institutions may move toward these technologies faster than they can address problems with data, processes, and governance.

He says generative AI can accelerate what an organization is already doing, but it "cannot compensate for weak data foundations, fragmented processes, or the absence of clear governance." He believes institutions achieving stronger results are those that treat AI as part of business transformation rather than as a standalone technology project.

He cites the experience of Ma'aden, saying its teams save more than 2,200 hours per month using Microsoft 365 Copilot, Copilot Studio, and Azure OpenAI Service. In his view, the speed of adoption should not be considered separately from the quality of the foundations supporting it. He says security and governance must advance at the same pace, so the question is not whether an institution is moving too quickly, but whether its organizational, data, and security infrastructure is moving with it.

From consuming technology to producing it

Al-Ghamdi believes the next phase in Saudi Arabia is not only about using global platforms, but also about increasing the country's ability to build technologies from within the Kingdom that can expand beyond it.

He says Microsoft's role in this area is to provide cloud infrastructure, AI and cybersecurity tools, data platforms, and support from its partner ecosystem to Saudi companies, startups, and developers. He also stresses that the local cloud region can help build and host solutions inside the Kingdom while meeting data residency and regulatory requirements.

He adds that the long-term opportunity is for more Saudi institutions to move "from consuming technology to producing it," including by developing intellectual property, platforms, and AI solutions capable of competing regionally and globally.

Digital sovereignty... and local capability

Asked about digital sovereignty and the distinction between data residency, operational control, and technological independence, Al-Ghamdi focused in his response on increasing local capacity to build and host solutions inside the Kingdom.

He noted that the presence of local cloud infrastructure enables innovators to develop and host solutions within Saudi Arabia while taking data residency and regulatory requirements into account, linking this to the shift from using technology to producing it.

Al-Ghamdi did not directly distinguish between data residency, operational control, and technological independence. Instead, he linked the issue to the ability of Saudi companies, startups, and developers to retain a greater share of technological value, knowledge, and intellectual property within the Kingdom.

A role beyond infrastructure

Al-Ghamdi traces Microsoft's role in Saudi Arabia to more than 25 years of work with the government, institutions, companies, developers, partners, and educational organizations, including cooperation with the Ministry of Communications and Information Technology in cloud, skills, and responsible AI.

He says the current phase raises the level of responsibility from simply providing technology to turning it into national outcomes, such as improving services and productivity, strengthening secure digital environments, and creating broader opportunities for Saudi talent.

Al-Ghamdi takes up his position after more than 15 years of experience at Microsoft. Before that, he led the company's public sector business in Saudi Arabia, covering government entities, national institutions, education, and healthcare, in addition to participating in initiatives related to government cloud regulation and the National Analytics Platform. His career has also included roles at Oracle and Google.

The three-year test

In response to a question about the criteria by which his tenure could be assessed after three years, Al-Ghamdi did not tie success to business volume or the number of products sold. Instead, he framed it more broadly around turning AI ambitions into tangible results.

"After three years, I will measure success through one question: Did we help Saudi Arabia turn its AI ambition into progress that people can see?" he says.

He defines this in terms of the expansion of trusted cloud infrastructure across institutions and critical sectors, the transition of AI from experimentation to measurable improvements in productivity, services, and customer experience, as well as an increase in the number of Saudi talents, partners, and developers capable of creating digital value from within the Kingdom.

Within this framework, the phase described by Al-Ghamdi centers on testing whether technology investments can move from adoption to execution, and from the use of models and platforms to measurable outcomes in productivity, services, skills, and intellectual property within the Saudi economy.


Xiaomi Sees Smartphone Cost Pressures Easing, Looks to EVs for Growth

Xiaomi is increasingly relying on electric vehicles and artificial intelligence as it seeks growth drivers beyond its increasingly saturated core business of smartphones. - File Photo
Xiaomi is increasingly relying on electric vehicles and artificial intelligence as it seeks growth drivers beyond its increasingly saturated core business of smartphones. - File Photo
TT

Xiaomi Sees Smartphone Cost Pressures Easing, Looks to EVs for Growth

Xiaomi is increasingly relying on electric vehicles and artificial intelligence as it seeks growth drivers beyond its increasingly saturated core business of smartphones. - File Photo
Xiaomi is increasingly relying on electric vehicles and artificial intelligence as it seeks growth drivers beyond its increasingly saturated core business of smartphones. - File Photo

China's Xiaomi Corp said the worst period of pressure on its smartphone business had passed as the pace of memory price increases looked set to slow in the second half, while it sees its fast-growing electric vehicle business delivering a larger share of revenue.

Xiaomi on Tuesday posted a 42.6% fall in second-quarter adjusted net profit to 6.2 billion yuan ($919.5 million), missing analysts' estimates, as historically high memory and other component costs squeezed margins for the maker of smartphones and electric vehicles.

Analysts had on average expected 6.6 billion yuan, according to LSEG data.

Revenue fell 6.1% from a year earlier to 108.9 billion yuan, also missing the 112.2 billion consensus forecast.

"Significant increases in key component costs, including memory, along with intensified industry competition, continued to create headwinds for our business," Xiaomi said in its earnings statement.

MEMORY COSTS REMAIN HIGH

In a post-earnings call, Xiaomi President William Lu said memory costs remained at historically high levels in the second quarter, as higher component costs weighed on margins in Xiaomi's smartphone and tablet businesses.

Xiaomi's smartphone revenue fell 7.5% year-on-year to 42.1 billion yuan, while its smartphone gross margin declined to 8.5% from 11.5% a year earlier.

Xiaomi, ranked as the world's No. 3 smartphone maker, shipped 31.2 million smartphone units in the quarter, down 26% from a year ago, for a second consecutive quarter of decline, research firm Omdia said.

With more than half its shipments priced below $200, Xiaomi was the most exposed among the top five smartphone vendors to memory cost inflation, Omdia added.

Yet Xiaomi said the pace of memory-price increases had started to slow and should continue to slow in the second half.

Lu said the most difficult period for the smartphone business had passed, adding that Xiaomi had adjusted its product mix and launch schedule.

EV BUSINESS PLAYS A BIGGER ROLE

Xiaomi is increasingly relying on electric vehicles and artificial intelligence as it seeks growth drivers beyond its increasingly saturated core business of smartphones.

Its EV, AI and other new initiatives segments accounted for about 23% of total revenue, up from 18.3% a year earlier.

EV revenue alone rose 15.9% to 23.9 billion yuan.

The domestic car market has been in steady decline since late 2025, while other Chinese carmakers are aggressively expanding exports. Xiaomi plans to enter European markets in 2027.

The loss from operations related to its EV, AI and other new initiatives was 2.6 billion yuan, reflecting the company's continued investments in those areas.

Xiaomi delivered 104,199 vehicles in the second quarter, up 28.2% from a year earlier.

In July, Xiaomi unveiled its SkyNomad SUV series, expanding beyond battery-powered sedans and crossovers into a category popularised by models from Chinese peers.