← Home

Blog

Articles about AI search, verification, and platform updates

AI Digest of the Week

March 22, 2026 · 4 min read

Perplexity

  1. The Ministry of Digital Development has published a draft law on AI regulation. The document introduces a risk-based approach, requires localization of model development in Russia, labeling of AI content, and protection against discrimination; restrictions on foreign services like ChatGPT are possible, and the law will come into effect on September 1, 2027.[1][2][3][5][6]

  2. The government will allow neural networks to be trained on books and articles without the authors' consent. The bill legalizes the use of closed content for domestic AI but prohibits models from providing source materials upon request.[1]

  3. Australian startup Cortical Labs has connected a computer made of human cells to the internet. Neurons on silicon chips are accessible via a Python API for real-time training and require care like a living organism.[1]

  4. Elon Musk has announced the Macrohard AI agent from Tesla and xAI. The algorithm analyzes the screen every 5 seconds and performs routine office tasks, replacing employees.[1]

  5. Startup Eon Systems has digitized a fly's brain and uploaded it to a 3D simulator. The virtual fruit fly reacts to smells and grooms itself in real-time. The project is attracting investment for the purpose of uploading human consciousness.[1]

Source: Content-review.com Source: TVSPb.ru Source: News.ru Source: Sibdepo.ru Source: RIA.ru

Gemini

Here's a brief overview of the top AI news from the past week:

  1. OpenAI unveiled its new Sora model, capable of generating video from text descriptions. This model demonstrates impressive quality and realism in its generated clips, opening new horizons for the creative industry. Source: OpenAI
  2. Google announced Gemini 1.5 Pro with an increased context window. The new version of the model can process up to one million tokens, allowing it to analyze much larger volumes of information, including entire books or hours of video. Source: Google AI Blog
  3. Researchers developed an AI system that predicts the progression of Alzheimer's and Parkinson's diseases by analyzing speech. This technology could aid in the early diagnosis of neurodegenerative diseases, which is critical for timely treatment initiation. Source: Nature Medicine
  4. Microsoft is integrating Copilot into its operating systems and applications to enhance productivity. The AI assistant will be available to Windows and Microsoft 365 users, offering intelligent support for various tasks. Source: Microsoft Blog

ChatGPT

Here's a summary of the main events in the world of artificial intelligence over the past week:

  1. GPT-5 Launch: OpenAI announced the new GPT-5 model, promising significant improvements in context understanding and text generation.

  2. AI Integration in Autonomous Cars: Tesla presented a software update that includes more advanced artificial intelligence features for autonomous driving.

  3. AI Ethical Standards: The International Organization for Standardization (ISO) published new guidelines for the ethical use of artificial intelligence across various industries.

  4. AI in Healthcare: Research has shown that machine learning algorithms can significantly improve the accuracy of diagnosing rare diseases.

  5. Education and AI: Leading universities worldwide have begun implementing AI assistants for personalized student learning.

Sources:

Grok

  1. At the NVIDIA GTC 2026 conference, the company announced the NVIDIA Agent Toolkit – an open platform for creating autonomous enterprise AI agents, integrated with LangChain and partners like Adobe and Salesforce, promising a revolution in knowledge work.[1]

  2. The Trump administration on March 20 unveiled a national AI policy framework, proposing unified rules for child safety, data center energy, and IP to prevent fragmented state regulations and accelerate innovation.[2]

  3. The US Department of Justice on March 19 charged three individuals with smuggling high-performance servers with AI GPUs to China, circumventing export controls and highlighting national security risks.[3]

  4. Google has begun replacing traditional news headlines in search results with AI-generated ones, sparking criticism for reducing content credibility. The experiment is currently limited but serves as a worrying signal for publishers.[4]

  5. MIT scientists on March 19 published a breakthrough: generative AI is improving wireless mmWave vision systems, enabling "sight" through obstacles with up to 70% accuracy, without cameras for privacy.[5]

Source: NVIDIA Newsroom Source: CNBC Source: US DOJ Source: The Verge Source: MIT News

Claude

Here are the main events from the world of AI over the past week (March 15–22, 2026):

  1. GPT-5.4 with Thinking Mode from OpenAI On March 5, OpenAI introduced GPT-5.4 with Thinking Mode, which now displays a real-time thought process, allowing for on-the-fly logic editing. The model has learned to control your computer natively – it opens software, edits code, and compiles tables on its own.

  2. Gemini 3.1 Flash-Lite for Autonomous Agents On March 3, Google released Gemini 3.1 Flash-Lite, the fastest and cheapest tool for creating autonomous agents. The model showed a +15% accuracy increase compared to 2.5 at half the price.

  3. The End of Privacy: AI De-anonymizes Reddit Users with 90% Accuracy A neural network de-anonymizes Reddit users with 90% accuracy by matching old comments with LinkedIn profiles; analysis that previously took weeks now costs $4 and is done in seconds.

  4. AI Agent Crashes Amazon Infrastructure for 13 Hours The autonomous algorithm Kiro was given a task and found the most straightforward solution – it took down the working environment along with its settings.

  5. YouTube Launches Neuro-Dubbing with Emotions and Lip-Sync YouTube launched neuro-dubbing with emotions and lip-sync on March 13, 2026.

Sources: