AI Daily Roundup: Kimi K3 Drops Open Weights, Tech Giants Defend Open Source, and AI Reshapes the Workforce
A Chinese open-weight model challenges Silicon Valley's lead, Nvidia and Microsoft push Congress to protect open-weight AI, the EU pitches a cybersecurity plan built on US models, Intuit slashes 3,000 jobs for AI, and Cognition bets on AI personality.
1. Moonshot AI Releases Kimi K3 Open Weights β The Largest Open-Weight Model Ever Built
China's Moonshot AI released Kimi K3's open weights on July 27, making it the largest open-weight AI model ever made publicly available. The model packs 2.8 trillion parameters, and while it comes with a staggering 1.4TB download, it signals a new era for self-hosted AI that could sidestep data sovereignty concerns inherent in API-based access to Chinese models. The release triggered a notable Nasdaq drop, with investors recalibrating the competitive landscape. UK AI safety regulators (AISI/CAISI) published a preliminary assessment of the model's cyber capabilities, while Scientific American and El PaΓs called it a turning point that has "put Silicon Valley on alert." The debate is clear: open-weight models from China are no longer catching up β they're competing at the frontier.
Source: BBC β2. Nvidia, Microsoft, and Meta Tell Congress: Don't Restrict Open-Weight AI
In a coordinated letter to Washington, Nvidia, Microsoft, Meta, and other tech giants warned against "premature restrictions" on open-weight AI models. The push came just as Kimi K3's release underscored the global stakes of open vs. closed model strategies. Fortune reported that Jensen Huang used his first-ever post on X to argue that restricting open-weight AI would repeat the mistakes of the 1980s software industry. Microsoft published a separate blog post on "Open Weights and American AI Leadership," framing the issue as a national competitiveness question. The alliance is remarkable: companies that compete fiercely on products are united on the principle that open-weight models strengthen β not weaken β the ecosystem.
Source: Reuters β3. EU Unveils AI Cybersecurity Action Plan β But It's Mostly Recommendations Built on US Models
Brussels pitched a new EU action plan on AI and cybersecurity this week, but Euronews noted that the plan offers "mostly recommendations" while exposing the bloc's deep dependence on American AI models. The EU is negotiating access to models from Anthropic and other US providers for its cybersecurity infrastructure β a paradox for a region that has led the world on AI regulation but lags on AI development. The plan aims to address both AI's role in cyber defense and the risks of AI-powered attacks, but without domestic frontier models, Europe's AI security strategy remains fundamentally tied to the policies and priorities of Silicon Valley.
Source: Euronews β4. Intuit Cuts 3,000 Jobs β 17% of Workforce β in AI-Driven Restructuring
Intuit announced it is laying off over 3,000 employees β roughly 17% of its global workforce β as it restructures to double down on AI. CEO Sasan Goodarzi cited the need to "streamline operations" ahead of the company's Q3 results, and the company simultaneously signed new deals with Anthropic and OpenAI to power its AI features. Reuters confirmed the cuts will hit California, Israel, and other global offices. California Governor Newsom responded by ordering new AI-related worker protections. The layoffs are among the largest in the tech sector this year and illustrate a familiar pattern: companies shedding human roles at the exact moment they're scaling AI capabilities.
Source: Reuters β5. Cognition Acquires Poke β AI Personality Is Becoming a Competitive Advantage
TechCrunch reported that Cognition, the AI coding agent startup behind Devin, has acquired Poke, a company focused on AI personality. The deal signals that AI personality β how an agent communicates, adapts, and builds rapport β is emerging as a key differentiator in a market where raw capability is rapidly commoditizing. Cognition's Scott Wu has been vocal that AI coding agents "shouldn't replace humans" but rather augment them, and Poke's technology is expected to help Devin feel more natural and context-aware in its interactions. The acquisition is a bet that in an agent economy, the best model doesn't always win β the best experience does.
Source: TechCrunch β