Nvidia wants to officially bring DLSS 5 to RTX 40 cards as well, where an unofficial mod version is already running.
Google has released WeatherNext 3. The AI model offers higher resolution, hourly updates, and is said to better capture dynamic weather patterns.
"Ulrich Wetzel – Das Strafgericht" is an institution on RTL; new episodes have now been produced with generative AI. They will be broadcast next week.
Google is advancing the integration of AI into everyday tasks. Gmail, Docs, and Keep can be controlled via natural language, also in German.
The Kubernetes-native AI platform Kubermatic AI aims to consolidate GPU resources and provide models as a shared service across teams.
Motorola is bringing the Edge 70 Plus, a new smartphone, to IFA 2026 in Berlin. Also launching are the Moto Watch Ultra and a special edition of the Razr 70.
Muse Spark 1.3 moves into the top group of current AI models and is particularly efficient in terms of performance and cost.
Nvidia buys Hugging Face for 12ドル.93 billion. The platform for open AI models will remain open, the company promises.
The digital association Bitkom sees a clouding of the IT job market with significantly fewer vacancies. However, there is still a shortage of skilled workers.
More victims of a school massacre are suing OpenAI. The company allegedly failed to inform authorities about the shooter’s violent intentions despite warnings.
A pilot project is starting on Monday at Dortmund’s main train station, where AI-supported video surveillance is to be tested.
Gemini 3.8 Flash is more powerful, but can cost more per task. Flash Cyber launches as part of a new Google security initiative.
The voice actor association VDS complains to the Federal Cartel Office about Netflix. Voice actors are forced to accept AI clauses for economic reasons.
AI makes even educated citizens insecure and must be democratized, demands Leif Lindner. The IFA CEO is enthusiastic about the Chinese development spirit
Norway’s Digital Minister considers facial recognition with smart glasses too much. An expert council is to develop proposals for stricter rules.
Neurodivergent talents are in demand in the tech industry. AI is intended to help autistic people and at the same time presents risks that could affect everyone
The US government supports OpenAI in its legal dispute with the New York Times. Training AI models is not copyright infringement and is essential for security.
A Bitkom survey shows: The majority of Germans favor data centers at their place of residence, but criticize energy consumption.
With OpenProject 17.8, AI assistants can now create, comment on, and link work packages – with the same rules as users.
Medical AI is being intensively researched, but patients' perspectives often fall by the wayside. A new study shows where research has gaps.
Anthropic offers enterprise clients Zero Data Retention again with Enterprise Frontier Safeguards – but they must take over abuse control themselves.
The flood of requests for new data centers in the USA is fueling concerns about a power bubble. Experts warn of ghost demand and rising costs.
Fable 5.1 leads in external benchmarks, while Mythos 5.1 is designed to handle more cyber tasks with loosened safety mechanisms.
AI agents execute malicious code in manipulated git repositories on startup – without user intervention, but with full rights.
Firefox 155 brings new AI features, a tracker counter, and improvements for containers – but causes problems in some corporate networks.
The Pentagon is now using its own versions of ChatGPT and Grok. What is the US military using the AI services for?
The Sonos Beam Ultra soundbar gets real, not just virtual, height channels, and the Ace Ultra headphones also play with music speakers, not just TV speakers.
BSI and Federal Network Agency form the nucleus for AISI Germany. "AI must remain controllable" is the motto for the new institute.
Google is expanding Gemini’s control functions under Android 17. With "Device Help", users should be able to adjust settings via AI dialogue.
AI scribes in the British healthcare system confuse medications and diagnoses. This technology is already in use in Germany.
OpenAI forecasts 1 billion dollars in annual revenue for its advertising business. This leaves 1.5 billion missing to reach the annual target of 2.5 billion.
Despite rising budgets, few companies are successfully scaling AI. Gartner shows the discrepancy between investment and broad adoption.
Users report that Anthropic has logged them out of Claude. This is to protect them from fraudsters.
Anthropic is to spend another 35ドル billion on AI computing power. Nvidia is securing the infrastructure, thereby strengthening demand for its own chips.
Zero-CVE images through Agentic AI: Red Hat’s Hummingbird project shows how AI agents automate container security – and where humans remain indispensable.
Nvidia and MediaTek deepen their long-standing partnership and expand their cooperation in AI data centers, PC chips, and the automotive industry.
Solid and precisely because of that successful: According to the survey, the open-source framework Django creates a stable and reliable basis for AI workflows.
Anthropic's "Model Hardware Standard" allows AI agents to directly control lab equipment, robot arms, and lasers.
For the first time, the EU Commission has classified an AI service as "very large", making it subject to stricter DSA rules, along with Roblox and Reddit.
The Debian project allows generative AI contributions. There is no obligation to disclose, transparency is merely recommended.