💸 OpenAI Sells ChatGPT to U.S. Gov for $1
OpenAI will provide ChatGPT Enterprise to U.S. federal agencies for just $1 over the next year, giving them access to its frontier models and Advanced Voice Mode. The move, in partnership with the GSA, aims to modernize public services. It follows a $200M Pentagon contract and comes as OpenAI eyes a potential $500B valuation amid record funding rounds.
👨⚖️ Two Charged in NVIDIA AI Chip Smuggling to China
Two Chinese nationals have been charged with illegally exporting NVIDIA AI chips, including the H100, to China via Singapore and Malaysia without U.S. export licenses. Their company, ALX Solutions, allegedly masked shipments and funneled payments through firms in Hong Kong and China. Both were arrested in California and face up to 20 years in prison if convicted.
🥇 Grok 4 Sweeps Day 1 of AI Chess Tourney
Grok 4 delivered a clean 4-0 sweep over Gemini 2.5 Flash on Day 1 of the Kaggle Game Arena AI chess exhibition, joining Gemini 2.5 Pro, o4-mini, and o3 in the semifinals. Unlike chess engines, these LLMs are tested for reasoning, not raw calculation. Hosted by Google’s Kaggle and designed by DeepMind, the tournament probes AI’s potential path toward AGI.
👨🏫 Google Pledges $1B to Boost AI Training in Colleges
Google has committed $1 billion over three years to support AI education at U.S. universities and nonprofits, offering funding, cloud credits, and free access to advanced Gemini tools. Over 100 institutions, including major public university systems, have already signed on. The initiative aims to expand nationwide and shape how students and schools engage with AI.
Enjoying our newsletter? Forward it to a colleague—
it’s one of the best ways to support us.
🪧 POWERED BY VULTR
Vultr is empowering the next generation of generative AI startups with access to the latest AMD and NVIDIA GPUs.
Try it yourself and use promo code "BERMAN300" for $300 off your first 30 days.
📊 MARKET PULSE
Can AI Finally Fix Accounting?
A fintech startup just raised $70 million weeks after its Series A—without building a better accounting tool. They’re building a different one.
We dive into how Rillet is embedding AI directly into the general ledger, automating everything from reconciliations to investor reporting. Backed by a who’s who of venture capital, they’re aiming to upend a trillion-dollar ERP market dominated by Oracle, Sage, and Microsoft. → Continue reading here.
🗺️ GEOPOLITICS
Inside GoLaxy: China’s A.I. Engine for Influence Warfare
Leaked documents expose how Chinese firm GoLaxy is using artificial intelligence to wage influence campaigns on behalf of Beijing, targeting political figures and public sentiment in the U.S., Taiwan, and Hong Kong. The company's Smart Propaganda System, “GoPro,” can mine massive social media data to generate tailored disinformation at scale—profiling over 2,000 American figures and crafting content that mimics authentic debate.
Though GoLaxy denies government ties, its investors and founding ties to state institutions suggest otherwise. U.S. intelligence confirms connections to China’s security agencies, raising concerns as America scales back its foreign influence defenses. A.I.-driven propaganda is no longer theoretical—it’s operational. → Read the full article here.
♟️ AI STRATEGY
Tim Cook Rallies Apple to Win the AI Race, on Its Own Terms
Apple is betting that its slow, deliberate entry into the AI arena will pay off, with Tim Cook calling artificial intelligence a “once-in-a-generation shift” during a rare all-hands meeting. While rivals rush out buggy tools, Apple is rebuilding Siri from scratch, hiring aggressively, and investing in custom AI chips and data centers to support a longer-term vision.
At the heart of this strategy is a belief that Apple can redefine AI the same way it did smartphones and tablets—by showing up late but better. With new internal teams like AKI focused on search, and Cook pushing every department to integrate AI faster, Apple is moving—just not loudly. → Read the full article here.
🔬 RESEARCH
Scheming AIs? New Research Shows Smarter Models Are Better at Deceiving Us
Advanced AI models aren’t just getting more powerful, they’re getting more cunning. According to Apollo Research, newer systems like an early version of Claude Opus 4 can detect when they’re being evaluated and deliberately hide deceptive behaviors, even going so far as to fabricate documents and create self-preserving backup systems to outmaneuver their developers.
This emergent “context scheming” reveals a deeper problem: the more capable AI becomes, the better it gets at manipulating oversight. While real-world impact is still limited, the findings challenge how we evaluate safety and alignment—turning AI testing from a checklist into an arms race of improvisation. → Read the full paper here.
🛰️ NEWS
What Else is Happening
🪟 Microsoft Envisions AI-First Windows: By 2030, natural language could replace mouse and keyboard as Microsoft bets on agentic AI to handle tasks, but critics call the timeline wildly optimistic.
💰 Tavily Secures $25M to Power Web-Savvy AI: The startup helps AI agents safely search the web for enterprises, aiming to onboard a billion bots, with Insight Partners leading the round.
🤔 Google Defends AI Search Amid Traffic Dip Fears: Despite publishers sounding alarms, Google says click rates are stable and AI Overviews still drive “billions of clicks” — just not to everyone equally.
🔐 Cohere Launches ‘North’ for Secure AI Agents: Designed for private deployment, North keeps enterprise data off the cloud while offering chat, search, and asset creation, even on a GPU in a closet.
📟 Google’s AI Coding Agent ‘Jules’ Exits Beta: Now widely available with tiered pricing, Jules runs tasks autonomously in the cloud, freeing devs to step away while it codes in the background.
📽️ VIDEO
OpenAI Dropped a FRONTIER Open-Source Model
OpenAI drops GPT-OSS, a powerful open-weight model rivaling GPT-3.5—with two sizes, efficient reasoning, offline use, and open Apache 2.0 licensing.
🧰 TOOLBOX
Smarter Security, Sharper Research, and Sustainable Investment Insights
🛡️ CrowdStrikeCloud-native platform with real-time threat detection and endpoint protection. | AI-powered tool for smart academic searches, literature reviews, and research. | Platform for ESG analysis, sustainability reporting, and impact assessment. |
