Capital, Agents, and Regulations: A Breakthrough Week in the Global AI Race

On February 2, SpaceX announced the acquisition of xAI^[1] , Elon Musk’s artificial intelligence company and creator of the Grok chatbot, in a stock swap deal exchanging 1 share of xAI for 0.1433 SpaceX shares. Together valued at 1.25 trillion dollars^[2] integrating rockets, Starlink satellites, the X platform^[3] and the Grok AI models, making it the most valuable private company in the world^[6]. After a January Series E round, xAI was valued at 250 billion dollars^[4], while SpaceX about 1 trillion dollars. Elon Musk announced an initial public offering around June 2026^[5] , aiming to raise close to 50 billion dollars while protecting SpaceX from potential legal liabilities linked to xAI, including an ongoing investigation into deepfakes generated by Grok^[40].

SpaceX–xAI Merger and Nvidia Investment

Simultaneously, Nvidia finalizes a 20 billion dollar investment^[7] in OpenAI, part of a broader funding round^[8] reaching 100 billion dollars and valuing OpenAI at roughly 830 billion dollars^[9]. OpenAI’s new models are trained on NVIDIA GB200 NVL72 systems^[10], further cementing OpenAI’s technological dependence^[11] on Nvidia’s architecture and raising the entry barrier for competing chip manufacturers^[41].

New AI Agent Models and Platforms

On February 5, OpenAI released the GPT-5.3-Codex model^[12], which merges the Codex and GPT-5 training stacks^[13] and operates 25% faster than its predecessor^[14]. The company classified its own model as having high capabilities^[15] in cybersecurity, recognizing its high capabilities in this area^[42], while admitting that earlier model versions helped debug and manage its training process^[43]. Due to abuse risks, OpenAI launched a trusted access program for security experts^[16], offering 10 million dollars in API credits for cyber-defense projects^[17] and temporarily delaying full API access^[18], making the model mainly available in ChatGPT for paying users^[44].

That same day, OpenAI also launched the Frontier platform^[19], aimed at enterprises to become an operating system for AI agents acting as digital coworkers. Frontier integrates with internal systems like CRM and HR databases^[20] and provides memory, agent onboarding, and detailed permissions management^[21]. Developed under the supervision of Fidji Simo, it launches without disclosed pricing^[22], already including Intuit, State Farm, Thermo Fisher, Uber, BBVA, Cisco, and T-Mobile^[46]. This is a direct response to the Claude Cowork offering from Anthropic, which on February 5 unveiled the Claude Opus 4.6 model^[23], with a context window of 1 million tokens^[47], targeting long-term professional tasks in finance^[24].

Agent Oversight Crisis and EU Regulations

The AI agents landscape is supplemented by a Gravitee report^[25], prepared with research firm Opinion Matters, published on February 3. Surveys of 750 IT directors in the US and UK estimated^[48] that organizations operate roughly around 3 million AI agents^[26], of which 53%, nearly 1.5 million, are neither actively monitored nor properly secured^[27]. As many as 88% of companies experienced a security incident involving an AI agent in the last 12 months^[28]. The deployment pace outstrips security teams’ capabilities^[49], and experts from Beauceron Security and Kapittx warn that deployment speed exceeds capacity^[29], while the global market urgently needs governance tools^[30] for agent systems.

The implementation of the EU AI Act regulation^[31] in Europe enters a critical phase. The European Commission was due by February 2, 2026, to prepare guidelines^[32] for the practical enforcement of Article 6 concerning the classification of high-risk systems, and full rules for such systems will take effect on August 2, 2026^[33]. Operators of high-risk AI systems must demonstrate comprehensive risk management^[34], while the schedule already includes a ban on unacceptable practices effective February 2025^[35]. Requirements for general-purpose models will apply starting August 2025^[36], meaning a need for rapid identification of own high-risk systems^[37] and full enforcement of the law^[38] for companies in Poland, alongside planning audits, documentation, and close cooperation with agencies such as the Office of Competition and Consumer Protection and the Personal Data Protection Office.

Related posts:

• Weekly AI Review: Gemini 3 Pro Dominance, $20B Megafunding for xAI, and New California Regulations
• Breakthrough Week in AI: China’s Offensive, Agent Platforms, and Security Crises
• NVIDIA, New Regulations, and a Wave of Funding: The Week AI Agents Enter the Mainstream
• Iran–Israel–USA Conflict: Status as of March 2, 2026