BBNews

Google introduced Gemini 3.8 Flash TTS and Flash-Lite TTS for creating and customizing voices, including control over accents and emotional delivery. The models are available through Google AI Studio, the Gemini API, and a range of Google products, expanding the tools available to developers and audio creators.

An Anthropic research team used Claude to analyze DNA data and found a new enzyme system with repeats resembling CRISPR. The company also announced a lab for fundamental research in biology; the results show how AI models can help generate hypotheses for subsequent experimental testing.

OpenAI introduced MentalHealthBench, a benchmark for evaluating the helpfulness and safety of AI responses in realistic conversations about mental health. Expert evaluation of these scenarios could help compare models and identify risks when they are used in sensitive areas.

Meta unveiled Muse Charm, a small wearable device designed for quick interactions with its Muse assistant. The company hopes to make the assistant available beyond glasses and smartphones.

Chicago’s mayor proposed a 12-month moratorium on new data centers while the city studies their effects on energy supplies, water, and residents. If approved, the proposal would temporarily limit infrastructure development in the city and could affect planning for new projects.

OpenAI outlined priorities and principles for third-party evaluators of its models. The publication includes no binding oversight mechanisms, so the principles alone do not guarantee that safety requirements will be met.

At a UN Security Council meeting, leaders of OpenAI, Anthropic, and Hugging Face called for international cooperation on the risks of autonomous AI systems. Their statement said that technology is advancing faster than safeguards are being developed, but no specific measures were announced at the meeting.

Leaders of OpenAI, Anthropic, and Hugging Face called on the UN to coordinate internationally and establish common standards for assessing AI risks. A US administration representative rejected new global regulatory mechanisms, saying they should not restrict the technology’s development.

The Ministry of Digital Development is considering a restriction under which only legal entities with direct contracts with mobile operators could sell SIM cards. The measure could change how telecom sales channels operate, but for now it is part of a package of proposals under discussion and has not been adopted.

The Federal Antimonopoly Service proposes banning energy companies not only from mining but also from providing infrastructure for it or owning related assets. If adopted, the amendments would affect energy market participants working with mining infrastructure.

Devices with verified Russian origins could be subject to a technology levy one-third lower than the standard rate. The measure is intended to encourage localization and inclusion in the official registry, but for now it is only a possible way of calculating the fee.

A US bill would allow rights holders to seek court orders blocking foreign piracy websites, including by involving VPN services and DNS resolvers. If adopted, it could expand the legal obligations of internet traffic intermediaries.

Gartner forecasts that global AI spending will rise 49.5% in 2026 to $2.7 trillion, followed by another 36.2% increase in 2027. Analysts note that some of the spending comes from relabeling existing IT budgets rather than reallocating them to AI.

SoftBank issued $10 billion in bonds and €1 billion in bonds. Some of the proceeds will go toward the final tranche of its $10 billion investment in OpenAI; once completed, SoftBank is expected to own about 13% of the company.

Growing public opposition in Texas to data center construction plans has become a subject of public and political debate. For operators and investors, this means that local community response can affect the progress of new projects.

An ASML representative said the company is not selling lithography equipment to European customers because no new chip fabs are being built in the region. This highlights the gap between Europe’s plans to develop its semiconductor industry and actual investment in new capacity.

McDonald’s 10-year plan includes developing its technology infrastructure and scaling its AI-powered ArchIQ ordering system. The company also plans to consolidate data in a global repository; the investment shows how large chains are linking automation and AI with operational modernization.

Sberbank increased the capital of an entity that owns a stake in the Element microelectronics holding company to 44.16 billion rubles. CNews sources link the funds to a possible consolidation of assets, but their specific purpose has not been confirmed.

Yadro plans to start mass-producing radio modules for LTE and 5G networks at its plant in Dubna in fall 2026. The 1.2 billion ruble project will combine two frequency bands in a single device, which could reduce the number of modules needed at base stations.

A group of hackers claims to have accessed an Oracle PeopleSoft system and stolen personal data belonging to FBI employees and their relatives. The claim of a large-scale breach has so far been made by the attackers; the incident highlights the risks for organizations using corporate HR systems.

Google outlined plans to add persistent memory across devices to Private AI Compute: data will be stored in encrypted form in the cloud, while the keys remain on the user’s devices. The architecture is intended to give AI assistants long-term context while keeping users in control of access to their data.

According to Netskope, the share of retail employees using unmanaged AI tools fell from 70% to 44%, while the share using company-approved solutions rose to 73%. At the same time, widespread use of built-in AI features makes it harder to control where customer data goes.

Google added tools to the open-source profiler XProf for cycle-accurate analysis of TPU kernels, including custom Pallas kernels. Developers will have an easier time finding bottlenecks in computing tasks that previously appeared in traces as a single opaque block.

According to a survey of UK executives, AI is changing hiring practices, and some companies have already reduced recruitment for entry-level roles. If routine tasks are handed over to AI, it may become harder for newcomers to gain the experience they used to build through this work.

A Visier study found that some employees feel pressure to use AI more often and fear job losses and fewer opportunities for advancement. The gap between corporate expectations and employees’ experiences makes training and clear rules for AI use increasingly important.

SUSE has offered employees with more than 10 years at the company the option to leave voluntarily or retire early. The company says the initiative reflects changes to its business structure; for experienced specialists, it signals possible job cuts.