NeuroVerdict requires JavaScript to run.

Blog

Articles about AI search, verification, and platform updates

AI News: The Day's Highlights

March 28, 2026 · 3 min read

Perplexity

SoftBank has issued a $40 billion loan for investment in OpenAI – this is the largest deal of the day in the AI world, highlighting how giants are raising the stakes in the race for leadership. Funds are moving away from standard venture capital injections towards large-scale financial structures to capture a share in a key sector.

This event signals the further dominance of AI in venture capital: money is flowing into proven models like OpenAI, rather than ambitious startups without monetization. Asia is also stepping up – South Korea's Rebellions has received

66 million in government support for AI chips.

Sources:

Startup and Venture Capital News March 28, 2026 (sergeytereshkin.ru)
AI on probation: what's wrong with regulation (dp.ru, March 26, 2026)

Gemini

Over the past 24 hours, one of the most significant news items in the AI world has been NVIDIA CEO Jensen Huang's statement that he believes AI has already reached the level of Artificial General Intelligence (AGI), meaning human-level intelligence. This statement has resonated widely within the community, as it could signify a transition to a new era in technological development.

Another important piece of news is OpenAI's closure of its AI video generation application Sora, which led to the termination of a billion deal with Disney. Although the reasons for this decision have not been disclosed, it raises questions about the future prospects of generative models for video.

It is also worth noting the ongoing discussions about AI regulation and the ethical risks associated with its application. For example, new research raises questions about how AI can be used as a therapist and identifies serious ethical risks. These discussions highlight the need for a responsible approach to the development and implementation of AI technologies.

Sources:

Tech.Onliner
ScienceDaily
Expert
Crescendo.ai
CBS News
FUTUREBY
ForkLog
South China Morning Post
Frank Media
NewsNow

ChatGPT

No significant news has been registered in the world of artificial intelligence in the last 24 hours.

Grok

Google has introduced a breakthrough algorithm called TurboQuant for AI, which reduces the memory consumption of machine learning models by 6–8 times. This allows for the optimization of data center operations, cutting energy costs by 50% or more, making AI more efficient without the need for vast amounts of expensive memory chips. The announcement has already caused a two-day drop in the stock prices of memory manufacturers like SK Hynix, exposing a rift in the AI market.

Sources:

Bloomberg: AI Breakthrough From Google Exposes Divide in Memory Chip Stocks (https://www.bloomberg.com/news/articles/2026-03-27/ai-breakthrough-from-google-exposes-divide-in-memory-stocks)
Mashable: Google AI compression technology saves data center energy (https://mashable.com/article/google-ai-compression)
VentureBeat: Google's new TurboQuant algorithm (https://venturebeat.com/infrastructure/googles-new-turboquant-algorithm-speeds-up-ai-memory-8x-cutting-costs-by-50)[[1]](https://www.bloomberg.com/news/articles/2026-03-27/ai-breakthrough-from-google-exposes-divide-in-memory-stocks)[[2]](https://mashable.com/article/google-ai-compression)

Claude

Looking for more specific news from the last 24 hours: Great, I've found up-to-date information on recent events. Here's the response:

This week, Nvidia CEO Jensen Huang stated that artificial intelligence has already reached the level of AGI (Artificial General Intelligence), sparking heated debate among researchers. Huang's statement has ignited intense discussions among scientists who believe humanity is still far from such a milestone; disagreements often boil down to the very definition of AGI.

Concurrently, three frontier models were released in March: GPT-5.4 (in three variants), Gemini 3.1 Ultra, and Grok 4.20. OpenAI quietly shut down its public Sora API due to unacceptable inference costs for video generation, highlighting economic limitations even for market leaders.

The Model Context Protocol (MCP) has surpassed 97 million installations, transitioning from an experimental status to fundamental infrastructure: all major AI providers now ship MCP-compatible tools. This signifies that AI agent systems are moving from demonstrations to real-world industrial applications.

Sources:

Follow us on socials

Subscribe where convenient: Telegram bot and channel or VKontakte community

@NeuroVerdictBot AI search in Telegram @neuroverdict_en News and articles VKontakte Community (RU)