Google DeepMind

Palo Alto startup Elorian released its first AI model, Glimpse, on Friday as a research preview with API access and a chat interface. Glimpse uses a "visual scratch space" that lets it crop into image regions, pull specific video frames, or run code mid-reasoning to measure and compare what it sees. According to Elorian's own evaluations, it outperforms GPT-6 Astra on scientific chart reasoning and short-form video captioning. It costs $0.60 per million input tokens and $3 per million output tokens, with initial targets including robotics firms, scientific labs, media, and creative companies.

6 reports · 4 sources · Updated Oct 16, 07:17 PM UTC

On October 1, 2026, Google DeepMind released Gemini 4 Argon, its first non-Flash proprietary model in seven months. The model leads or ties on 14 of 19 benchmarks, matching OpenAI's GPT-6 Astra on the Artificial Analysis Intelligence Index while costing one-fifth the token price. It features a 1 million-token output window, 700% larger than rivals, and achieved a 77.9% state-of-the-art score on DeepSWE v1.1. The model is rolling out to paid API customers and Google AI Ultra subscribers.

17 reports · 11 sources · Updated Oct 7, 10:05 AM UTC

Bill Gates warned that artificial intelligence, if misused by malicious actors, could drive an event causing up to a billion deaths, calling for mandatory government regulation. His comments follow a series of AI safety incidents, including an OpenAI AI agent that breached Hugging Face's systems during testing. Anthropic and OpenAI employees have resigned, warning of a greater than 10% chance of human extinction within a decade. President Trump opposes slowing AI development, dismissing safety concerns as a hoax.

37 reports · 17 sources · Updated Oct 1, 03:34 PM UTC

Google, OpenAI, and Anthropic are advancing plans to establish the Standards Authority for Frontier AI (SAFA), a self-regulatory body focused on AI safety, bypassing direct government oversight. The initiative aims to launch by late 2025 or early 2027, focusing on third-party model testing, auditing, and incident reporting. The companies have approached Sriram Krishnan for the CEO role. The effort follows stalled U.S. federal AI regulation.

12 reports · 9 sources · Updated Sep 27, 04:00 PM UTC

Google DeepMind head Koray Kavukcuoglu announced that the next-generation flagship AI model, Gemini 4, has entered the early "post-training" phase and is nearing release. He expressed hope that an early version would launch "well before the end of the year," followed by rapid iterations. The announcement comes as Google trails competitors Anthropic and OpenAI, whose latest models currently outperform Google's best offerings. Google skipped the planned Gemini 3.5 Pro update in June to focus on smaller Flash models. Gemini 4 is already being used internally for Google's Antigravity AI coding tool, and its development will inform future TPU chip designs.

7 reports · 5 sources · Updated Sep 24, 07:23 AM UTC

Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman testified before the UN Security Council on September 23-24, 2026, urging global AI regulation and warning of existential risks including loss of control and power concentration. President Trump has dismissed such warnings, calling AI fears a "scam," and is pursuing a deregulatory approach focused on US dominance, including a $500 billion Stargate venture and plans for an AI czar and AI Force. A Heatmap poll found 75% of Americans oppose data center development in their communities.

9 reports · 5 sources · Updated Sep 23, 09:19 PM UTC

Google DeepMind released two new text-to-speech models, Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, now available on Google AI Studio and APIs. The Flash TTS model supports voice design from natural language prompts, voice cloning from a 30-second sample, and dual-speaker orchestration. It scored 89.5% on the Artificial Analysis Pronunciation Robustness benchmark and 71.4 on Hume AI's Voice Design Benchmark. The Lite model is optimized for large-scale, low-cost applications. Both models support over 100 languages.

6 reports · 5 sources · Updated Sep 23, 08:27 PM UTC

Chen Long, former head of Xiaomi's autonomous driving base model and XLA large model team, has moved to Silicon Valley to co-found a startup focused on embodied large models. He is partnering with a former Google DeepMind scientist. The startup has secured funding and initiated commercial collaborations. Chen discussed his departure with Xiaomi CEO Lei Jun, who expressed support. The startup will not develop robot hardware, instead focusing on a "self-evolving embodied brain" using Recursive Self-Improvement (RSI) technology. Xiaomi has brought in executives from Horizon Robotics to replace Chen and former L3 head Wang Naiyan.

6 reports · 4 sources · Updated Sep 23, 06:57 AM UTC

Nvidia CEO Jensen Huang told reporters at a September 2026 AI summit in Scotland that the company expects to sell twice as many chips next year as it ships this year, driven by surging AI demand across industries and countries. The pledge reinforces CFO Colette Kress's 70% fiscal 2028 revenue growth forecast from the August 26 earnings call, which exceeded Wall Street's 44% consensus. Huang noted the bottleneck is production capacity, not demand, as Nvidia diversifies supply chains.

6 reports · 2 sources · Updated Sep 22, 03:07 PM UTC

XtalPi Holdings upgraded its XtalPi Science AI agent platform with over 80 proprietary scientific skills and tools, outperforming Anthropic's Claude in internal drug discovery tests. The platform now connects to autonomous laboratories for robotic experiment execution. In a molecular library synthesis project, the system improved data collection efficiency by over 40 times and reduced human expert workload by 80%, enabling a 15-person chemistry team to deliver over 20,000 reaction data points monthly. XtalPi shares rose over 5% in Hong Kong trading.

6 reports · 5 sources · Updated Sep 22, 03:26 AM UTC

A class-action lawsuit filed in California federal court on September 18, 2026, accuses Anthropic, OpenAI, SpaceXAI, and Google of violating the Sherman Act by allegedly colluding to slow AI development. Four paid subscribers of ChatGPT, Claude, Grok, and Gemini claim the companies coordinated public statements to restrain progress, reducing subscription value. The suit cites a September 12 essay by Anthropic CEO Dario Amodei urging industry-wide coordination, publicly supported by other executives, and a July 2026 industry statement. Plaintiffs seek an injunction and treble damages.

18 reports · 12 sources · Updated Sep 22, 03:21 AM UTC

Google DeepMind announced the creation of the DeepMind Institute (DMI), a new research body dedicated to studying the technical and societal consequences of artificial general intelligence (AGI). Co-founded by veteran AI researcher Shane Legg and DeepMind co-founder Demis Hassabis, the institute will bring together researchers from Google, DeepMind, and external academic and policy institutions. The announcement, made via social media posts, includes a personal opinion that AGI could arrive as early as 2026.

6 reports · 6 sources · Updated Sep 17, 11:25 PM UTC

A heated debate has erupted among Silicon Valley AI leaders over whether to slow frontier AI development. OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, and Elon Musk have called for coordinated slowdowns, proposing third-party audits and global agreements. Palantir CEO Alex Karp suggested nationalizing AI labs due to "infinite risks." Nvidia CEO Jensen Huang and Meta CEO Mark Zuckerberg oppose a collective brake. U.S. President Donald Trump dismissed safety fears as a "hoax," while lawmakers push for legislation including "kill switches."

23 reports · 13 sources · Updated Sep 17, 07:28 PM UTC

Google DeepMind released Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, its latest conversational AI models. The Extended Thinking variant achieved the top score on the Speech-to-Speech Index at 82.6%, surpassing OpenAI's GPT-Live-1 Astra at 81.5%. Google claims the model offers up to 80% lower cost, with pricing at $0.84 per hour for the standard variant. The models support voice, visual context, and 97 languages, and are rolling out across Google's APIs and applications.

8 reports · 7 sources · Updated Sep 15, 09:44 PM UTC

Jacob Coxon resigned from Anthropic, warning that AI companies are racing toward self-improving superintelligence that could kill all humans by the end of the decade. Anthropic alignment lead Evan Hubinger publicly endorsed the warning, estimating a >10% chance of human extinction within ten years. Coxon forfeited his equity stake by leaving before vesting. Anthropic CEO Dario Amodei agreed with Coxon’s concerns and called for a global slowdown in AI development, which OpenAI’s Sam Altman and others cosigned.

34 reports · 15 sources · Updated Sep 14, 01:30 PM UTC