作者 Timur Sırt
地点 Istanbul
Artificial intelligence is advancing much faster than expected, but the systems needed to monitor, oversee and govern the development are failing to keep pace
Competition and innovation in AI technology are moving at extraordinary speed. But the technologies designed to monitor and govern its development have fallen behind. In other words, AI has become a rapidly expanding force without sufficient oversight. What is more, that power is becoming concentrated in companies in a handful of countrie
展开查看正文
Artificial intelligence is advancing much faster than expected, but the systems needed to monitor, oversee and govern the development are failing to keep pace
Competition and innovation in AI technology are moving at extraordinary speed. But the technologies designed to monitor and govern its development have fallen behind. In other words, AI has become a rapidly expanding force without sufficient oversight. What is more, that power is becoming concentrated in companies in a handful of countries.
According to the 2026 AI Index Report, published by Stanford University's Institute for Human-Centered Artificial Intelligence (HAI), AI grew and advanced far faster than expected in 2025. One of the report's most striking findings is that the performance gap between U.S. and Chinese models has effectively closed.
After DeepSeek-R1 briefly matched the best U.S. model in February 2025, Anthropic's model was only 2.7% ahead of its rivals as of March 2026. The U.S., meanwhile, remains the country producing the largest number of top-tier models, while China leads in publication volume, citations and patent output.
Winning Mathematical Olympiad
The capabilities of these models also paint a "lumpy" picture.
Google DeepMind's Gemini Deep Think model has demonstrated performance at a level capable of winning a gold medal at the International Mathematical Olympiad, while the same systems can correctly read an analog clock only about 50% of the time.
A similar contradiction can be seen in robotics: robots that achieve success rates of up to 89% in controlled laboratory environments can complete only 12% of everyday tasks in real homes.
Revenue keeps rising
The report also devotes considerable attention to the technology's social and economic impact.
Generative AI, or GenAI, has spread faster than personal computers and the internet, reaching a global user adoption rate of 53% within three years. Enterprise adoption has reached 88%, while the estimated annual value generated by generative AI tools for U.S. consumers has reached $172 billion.
But this growth also comes at a cost: employment among younger workers aged 22 to 25, particularly in areas such as software development, is showing a marked decline, even though the clearest productivity gains in the same sector are being recorded in this field.
The U.S. is far ahead on the investment front. Private AI investment in the U.S. reached $285.9 billion in 2025, 23 times the level in China. Yet the U.S. is losing some of its ability to attract global talent. The number of AI researchers and developers moving to the country has fallen 89% since 2017, with the decline reaching 80% in the past year alone.
Türk Telekom launches leadership initiative with Stanford
One of Türkiye's leading telecoms and technology companies, Türk Telekom, has launched an education partnership with Stanford Center for Global and Online Education (CGOE) as part of the Türk Telekom Academy Leadership School.
The "Shaping Culture, Leading Change" program aims to strengthen the strategic capabilities of senior executives to lead corporate transformation.
Iskender Bayrak, deputy general manager in charge of human resources at Türk Telekom, said they see bringing together the power of technology with their highly skilled workforce as one of the most important elements of their digital transformation journey.
"We attach great importance to ensuring that our managers create a working environment that gives their teams confidence, supports development and makes every employee feel valued. Through this program, we will make a strong contribution to their development by taking our leaders' capabilities in innovation, change management and strategic communication to the highest level," Bayrak said.
Led by Türk Telekom Academy and launched in partnership with Stanford Center for Global and Online Education, the "Shaping Culture, Leading Change" program aims to strengthen the strategic capabilities of senior executives to lead corporate transformation.
Complead becomes compliance, measurement partner for institutions in AI era
Reaching more than 800 organizations across more than 80 countries in six years, Sanction Scanner has rebranded as "Complead" and entered the U.S. market with a new New York-based entity.
Founded in Istanbul six years ago, Sanction Scanner has unveiled its new identity as it evolves into an AI-powered compliance, risk and fraud infrastructure provider. The company will now operate under the Complead name. The Türkiye-based startup currently serves more than 800 corporate customers in over 80 countries.
Compliance infrastructure for global giants
Alongside leading Turkish institutions such as Türk Telekom, iyzico, Yemeksepeti, Togg, Midas and Hepsiburada, the company provides compliance infrastructure for organizations including NN, Unity, Zurich, Loomis, Stellantis, Airhelp and the United Nations.
What began as a sanctions-list screening service has evolved into an integrated platform covering everything from customer risk profiling and transaction monitoring to fraud detection and ultimate beneficial ownership analysis.
"The name Sanction Scanner was right on the day we founded the company because what we did was simply scan sanctions lists. The name was a snapshot of our company, and our company has grown beyond that snapshot," Fatih Coşkun, founder and CEO of Complead, said.
Compliance in AI era
At a time when AI is becoming increasingly widespread across finance and compliance, Complead is placing an emphasis on making decisions auditable.
The company's AI agents and Fusion architecture automatically resolve 77% of routine compliance cases. The system, which can reduce false-positive rates by up to 97%, does not produce decisions in audit processes whose sources cannot be identified or explained.
According to Coşkun, compliance teams are caught between pressure from boards to improve efficiency and regulators' expectations for explainability. "We are building precisely at the intersection of these two needs," he said.
Next stop: U.S.
Following its global structure based in London, Complead has formalized its presence in the U.S. market with a new legal entity and local team in New York, while already gaining experience in the market through American customers such as APMEX, Hometap and Candex.
Coşkun emphasized that the U.S. accounts for the world's largest financial crime compliance spending, while the systems dominating the market are cumbersome and expensive solutions built decades ago. Services for existing customers will continue without interruption, contract changes or infrastructure migration.
"Our goal is not to be a good alternative; it is to be one of the companies redefining how financial crime is fought around the world," Coşkun said.
Smart glasses: Your personal assistant or shopkeeper watching you?
Meta VR Glasses' eye- and hand-tracking-focused design is putting data privacy back in the spotlight. Living behind smart lenses may feel like living in a glass house or turning into a customer watched by a shopkeeper at all times.
At Connect 2026, Meta introduced its new Meta VR Glasses, weighing around 100 grams. According to the company, the device packs a cinema, a stadium seat, a workspace and a gaming console into a pair of glasses. Its most striking feature is the use of the eyes and hands as the primary inputs.
Without a controller, users can start a movie, select an app or make a search. From a usability perspective, that is impressive. From a privacy perspective, however, it raises a question. With the SDKs already open, how applications will access this data is being designed from day one.
How an app that uses gaze data "only for the interface" will be presented to users compared with one that collects the same data for analytics could become one of the industry's biggest debates in the years ahead.
Shopkeeper who never stops watching
Imagine walking around a store without saying a word to anyone. But what if the shopkeeper were recording how many seconds you looked at each shelf, how often your eyes landed on a product's price, when you hesitated and which products excited you?
With devices operated by controllers, the information was limited to the buttons you pressed; in other words, only what you did. With eye and hand tracking, your gaze itself becomes an input. That means moving from asking, "What did you click?" to asking, "What did you look at, for how long and when did you react?" And all this could happen without you knowing exactly when the data is being captured.
Meta VR Glasses open the door to entering a cinema without holding a controller and creating a workspace in midair. But the price of that convenience is trust in the glasses watching you. Living in a glass house may sound appealing if the view is beautiful; we just need to know who gets to pull the curtains and who doesn't.
