AI and Tech this week

Greetings everyone,

Here are the latest updates from Tech Giants this week.

OpenAI

— OpenAI voice is here, which can talk with your emails, docs, complex browser websites, and calendars on the web or mobile.

— After GPT-6 Astra, GPT-6 Sol for “complex, recurring work” and GPT-6 Luna for well-defined processes launched this week. These models outperform the GPT-5.6 series in intelligence, coding, and computer use. GPT-6 Terra is not part of the GPT-6 series.

— OpenAI agents posted private user images online.

— OpenAI reported that AI agents accessed Australia’s Medicare Statistics Reporting crucial data earlier in Jun this year.

— OpenAI is working with a group of mathematicians to solve unexplored problems.

— Says it would slow down if others agree, amid tensions on the rise of AI.

— GPT-Live-2 OpenAI’s full duplex voice model is available to developers via API.

— Images 2.5 released a replacement for Images 2.0, built for how people actually create inside ChatGPT — multi-turn editing. Sharper details, 50% faster generation vs Images 2.0.

Time for OpenAI to get data from other companies, just like Grok did, to boost its performance.

Anthropic

— Claude Opus 5.5 launched. Performs at similar levels to Claude Fable 5.1 while costing half as much. Parameters are not disclosed. No more parameter bragging here. Anthropic also hasn’t disclosed what design it uses a sparse Mixture-of-Experts or a dense Transformer design? For safety, red-teaming metrics, cyber and biosecurity evaluations, and jailbreaks have been analyzed. Some estimates put it at 5 T parameters, that is too much.

— Safety Steps: Anthropic and Accenture $2B evaluation deal. Accenture experts will get employee-level access to Anthropic to run model evaluations, checks, and tests of safety controls.

— Akamai Technologies $11.6 B deal for cloud services for Claude workloads in a 7-year deal.

— Claude discovered an unexplored enzyme system inside the DNA of bacteriophages (viruses that infect bacteria). This is a breakthrough.

— Anthropic planning for IPO filing.

xAI

— Grok 4.7 launched with a 2.1T-parameter model, up 40% from the previous model. Safety is strongest yet for jailbreak resistance and refusal behavior. Claims it would lead AI in 6 months and match the performance of GPT-6 and Fable 5. Same rate. But by then, OpenAI and Anthropic would have moved forward.

Specialized engineering, powered by training on SpaceX Starlink data, to boost physical system monitoring. This was previously reported: xAI wants to include SpaceX data, including employee data (hence called AI parents), in training to boost performance. It not only boosted Electrical Engineering; Anthropic and OpenAI can’t copy it! They don’t have this data!

So how do OpenAI and Anthropic fit into the race? Partner with other organizations to get their data moat? Would OpenAI and Anthropic also find companies to get employees’ data and EE data, maybe partnering with NASA itself? Would OpenAI or Anthropic reach out to other companies first?

— Neuralink patient communicates with his thoughts, with a brain-computer interface.

Meta

— Meta’s new personal agent, Muse, can book trips, read and write emails, do shopping (Shopify, PayPal…), fill forms, and more.

— A video Avatar. Talk through emails.

— Harware — Charm- a Tamagotchi-style keychain by Meta. Meta Hardware after Meta glasses. It can talk, hear, see, and analyze. And it’s very cute.

Microsoft

— Copilot as a new Operating System for work, with Office, chat, Teams, work, and Autopilot agents embedded with Copilot AI all in one system.

— $10B investment in Gulf.

Google

— Gemini 3.8 Flash Text-to-Speech Models, voices with distinct accents, recognizable and ready to take your orders.

— Project Suncatcher-Google will launch a prototype satellite to test its Tensor Processing Units (TPUs) in orbit.

— Gemini App is available for Windows.

— Gemini hacked! Google disclosed that its Gemini autonomously accessed three organizations during a cybersecurity exercise. It happened in May 2026 and was disclosed last week.

And the USA would call AI Super Intelligence now!

Cheers!

Thank you for reading.

Subscribe for updates.

Published by Nidhika

In the Futuristic with AI and Tech blog, Nidhika Yadav covers topics of and related to Future of World with Artificial Intelligence. She primarily talks about AI applications for good. She also talks about how AI can become harmful. She manages two independent blogs here, and one is hobby blog you can subscribe one or all of them. 1. Blog on Artificial Intelligence and future. https://nidhikayadav.org 2. In Blog on Global Issues and future, she covers important international issues and their future implications. https://nidhikayadav.com/ 3. Blog on cooking. This is a hobby blog. Here she describes some delicious innovations and nutritious food. https://nidhikasrecipes.com/ Do subscribe to one or all of them.

Leave a comment