Perplexity
The Ministry of Digital Development of Russia has published a draft federal law on the regulation of artificial intelligence, which introduces a framework for regulating the industry, defines the rights and obligations of market participants, and mandates the labeling of AI-generated content. The document requires that all stages of AI model development and training be conducted exclusively within the Russian Federation by Russian citizens or legal entities, ensuring technological sovereignty and protection of data from leaks. This could limit or prohibit foreign services like ChatGPT, Claude, and Gemini, with an effective date no earlier than September 1, 2027.[3][4][5][6]
The bill has been put forward for public discussion and has become one of the most debated topics, highlighting a strategic shift in the country's approach to AI. It solidifies the definition of AI, obliges large platforms to store user data in Russia, and strengthens control over cross-border technologies to prevent manipulation.[3][5][6]
Source: Ministry of Digital Development of the Russian Federation, Draft Law Source: Lenta.ru Source: Sibdepo.ru
Gemini
Alibaba has introduced Qwen 3.5, an updated version of its flagship model focused on multimodal tasks and agent-based operations. The new model can analyze videos up to two hours long and is optimized to run on consumer hardware without the need for massive GPU clusters.
Progress is also being made in robotics: a new tomato-picking robot has learned to predict the ease of fruit harvesting and adapt its approach, increasing success rates to 81%.
Additionally, researchers from Swansea University have discovered that artificial intelligence can act as a creative partner for humans, enhancing creativity in design tasks.
ChatGPT
Here's the most significant AI news from the past 24 hours:
OpenAI has announced a new target: to create an autonomous AI-powered "research intern" system. This system is intended to independently handle a limited number of specific research tasks by September 2026. It will serve as a starting point for a more complex multi-agent research platform planned for implementation by 2028 (reddit.com).
Grok
The White House yesterday unveiled its long-awaited framework for federal AI regulation, urging Congress to adopt minimally burdensome rules and preempt state laws to avoid a "patchwork" of regulations that hinder innovation.[1][2] This is a key step by the Trump administration to maintain U.S. leadership in the global AI race with China, where fragmented regulation threatens the economy and national security.
The document proposes six priorities: tools for parents to control AI-generated child content, expedited permits for data centers with on-site energy generation, enhanced efforts against AI-powered fraud, a balance of copyright for model training, and a ban on government pressure on AI companies for censorship.[3] Reactions are mixed – tech investors praise the freedom for innovation, while critics fault the lack of accountability for risks.
Source: CNN Business Source: TechCrunch Source: NBC New York
Claude
OpenAI has released GPT-5.4 with a record-breaking context window of 1 million tokens and the ability to independently perform multi-step tasks. On the OSWorld-V test, the model achieved 75% accuracy, surpassing the human result of 72.4%, symbolizing a shift from AI as a chat assistant to AI as an autonomous digital colleague.
This is a significant breakthrough because it's not just an improvement to a language model, but a qualitative leap in its capabilities. The release is scheduled for March 2026, coinciding with iOS 26.4. Simultaneously, other significant shifts are occurring: OpenAI has surpassed $25 billion in annual revenue and is considering an IPO for late 2026.