What events in the field of AI took place during the week

Last week, major technology companies introduced updates in the field of artificial intelligence (AI): Google released a new audio model and expanded voice search to more than 200 countries, OpenAI opened access to the reasoning model for free users, and Anthropic was at the center of a source code leak scandal. All the details are in the Izvestia article.

Google has released a new audio model Gemini 3.1 Flash Live

Google has introduced Gemini 3.1 Flash Live, an updated audio model for real—time conversations. According to the developers, this is the company's highest-quality voice model to date: it reacts faster to cues, reproduces intonation more accurately, and works more steadily with fuzzy speech. Access for developers is available through Google AI Studio and Gemini API.

"We are expanding Gemini's capabilities to create real—time dialogues," the company's official blog says.

The model is already integrated into Search Live and Gemini Live and is available in more than 200 markets.

Prompta and contra: how Microsoft influences the outcome of the trial on the ethics of using AI

Leading neural networks analyzed the chances of concluding a deal between the Pentagon and the company Anthropic

OpenAI has opened GPT-5.4 mini for free users

OpenAI has opened access to the GPT-5.4 mini model in ChatGPT. It is available to free users and subscribers of the Go tariff through the Thinking function in the "+" menu. For paid Plus and Pro subscribers, the model is activated as a backup when the GPT-5.4 Thinking limits are exceeded during periods of high load. The GPT-5.4 mini is optimized for speed and lower computing costs while maintaining basic reasoning capabilities. The previous GPT-5 Thinking mini model will be disabled for 30 days.

Google Search Live has been launched in more than 200 countries

Google has expanded the Search Live feature to more than 200 countries and territories where AI mode is available in search. The tool allows you to conduct a multi—way dialogue with the search engine in voice format, ask clarifying questions and receive contextual answers in real time, including using the device's camera.

According to the company, the Search Live extension was part of a massive Gemini update in March aimed at turning users' devices into "proactive assistants" for work and everyday tasks. Previously, the feature worked only in a limited number of regions. AI-enabled voice search is now available worldwide wherever AI Mode is activated.

A gold mine: how artificial intelligence is changing the economy of scammers

Interpol announced a 4.5-fold increase in the income of intruders due to neural networks

Anthropic has blocked thousands of other people's repositories on GitHub

In an attempt to remove the leaked Claude Code source code from public access, Anthropic sent massive DMCA complaints to GitHub, which affected thousands of third-party repositories. The company admitted that blocking other people's projects also turned out to be a mistake.

This was the continuation of a series of failures: on March 31, a debugging file with the full source code Claude Code was discovered in the npm registry - about 512 thousand lines of TypeScript. Before Anthropic reacted, the code repository on GitHub had been saved more than 41,500 times. A few days earlier, about 3,000 internal documents of the company were publicly available, including a draft of the announcement of the new Claude Mythos model, Fortune reported.

All important news is on the Izvestia channel in the MAX messenger.