AI News Digest, July 7: Open-Weight AI Just Pulled In $800 Million

Open weight model funding just had its loudest week of the year. A cloud built for open-source models raised $800 million, and its founders say inference bookings now top a billion dollars a year. Meanwhile, the enterprise-software giants are racing to make AI agents safe enough to trust with HR and finance data. And in Geneva, 193 governments sat down to argue about who governs any of this. Three threads, one knot. The money, the guardrails, and the rules are all moving at once. Here is what changed, and what your team should actually do about it.

Open-Weight Model Funding Just Hit a New High

Together AI closed an $800 million Series C on July 1. The round values the company at $8.3 billion. Aramco Ventures led it, with Nvidia, General Catalyst, Vista Equity, and Salesforce Ventures joining. (Source: TechCrunch)

What the round actually says

Together AI runs a full-stack cloud for open-weight models. That means training, fine-tuning, and inference for models you can download and self-host, rather than closed APIs. The company says its inference stack runs about twice as fast at up to 60% lower cost. It also says annual bookings now top $1.15 billion. (Source: Together AI)

For context, its last round was $305 million at a $3.3 billion valuation, roughly 16 months ago. So the valuation more than doubled in a year and a half. On top of the equity, Together lined up commitments for over 500 megawatts of compute, funded separately. (Source: SiliconANGLE)

Why open-weight model funding matters for founders and HR leaders

Here is the shift underneath the headline. Open-weight models used to be the cheaper-but-worse option. That gap is closing fast. When a specialist cloud runs open models faster and cheaper than closed ones, the math for buyers changes. However, this is not a reason to rush anything.

For a 50 to 500-person company, the practical takeaway is narrow. Most of you will never touch raw inference. But your HR and ops vendors will, and their cost curve is bending down. As a result, expect the AI features inside your HR tools to get cheaper or better this year, because the compute under them just got cheaper. So ask your vendors a blunt question. Which models run your AI features, and did your price move when their costs did? Our guide to AI agents for HR is a good place to see where those features are heading.

What to do this week: if a vendor is about to raise prices on an AI add-on, push back. The cost of running these models is falling, not rising. Use that fact in the negotiation.

The Real HR Question Is Not Which Agent, But Whether You Can Trust It

On June 2, Workday shipped a set of tools for building and vetting AI agents that touch HR and finance data. The headline piece is Agent Passport. It gives each agent a set of digital stamps that show which security and compliance tests it passed, who verified them, and against which standards. (Source: PR Newswire)

Specifically, Agent Passport checks agents against public frameworks like the OWASP LLM Top 10, the NIST AI Risk Management Framework, and MITRE ATLAS. Cisco is the first outside partner issuing those stamps. Early access opens in the second half of 2026, with general availability targeted by year-end.

So what for your team? As HR adds agents for onboarding, payroll, and performance, the risk shifts. It moves from “does it work” to “can I prove it is safe with employee data.” That is a governance problem, not a features problem. Before you let any agent read payroll records, ask how you would audit it. If you cannot answer, you are not ready to deploy it. Closing the AI skills gap in HR is part of the same job.

Governments Are Now in the Room, Too

On July 6 and 7, the UN convened its first Global Dialogue on AI Governance in Geneva. All 193 member states were invited, alongside industry, civil society, and researchers. The goal is to build common approaches to AI governance so the rules reflect every country, not just the biggest labs. (Source: United Nations) The talks are non-binding, though, so do not expect new rules overnight.

The binding deadline is closer to home. On August 2, 2026, the EU AI Act’s obligations for general-purpose AI models become enforceable, and the Commission gains fining powers of up to 15 million euros or 3% of global turnover. (Source: EU AI Act) If you employ people in the EU or sell there, this is a real date. So ask whether your HR and payroll tools rely on general-purpose models, and whether your vendors are ready for it.

Quick Hits

• OpenAI previews a cheaper model family. OpenAI began a limited preview of GPT-5.6, a three-tier lineup named Sol, Terra, and Luna, with lower-cost mid and entry options. Access is limited to about 20 partner organizations first, after OpenAI shared the models with the US government, with a wider release expected in the coming weeks. (Source: OpenAI)
• An agent that lives in your keyboard. Singapore’s Acti launched an agentic keyboard for iOS and Android. It does not just suggest words. It can take actions inside the apps you already use. (Source: TechCrunch)
• Defense AI pulls a mega-round. German drone maker Quantum Systems signed a $1.2 billion Series D at about an $8 billion valuation, co-led by Blackstone, Airbus, and others. Its drones logged more than 19,000 missions in Ukraine last year. (Source: Quantum Systems)

The through-line this week is simple. Cost and control are moving in opposite directions. Open-weight model funding is pushing the price of AI down, while regulators and governance frameworks are raising the bar on how you deploy it. For a lean team, the winning move is boring. Get your people data clean, pick vendors who can prove their AI is safe, and automate only the workflows you understand. Asanify’s global HRMS platform and its guide to AI payroll automation are a practical place to start.

Open-Weight Model Funding: Your Questions Answered

Why is open-weight model funding suddenly a big deal?
Because the cost gap with closed models is closing. Together AI raised $800 million at an $8.3 billion valuation and says its open-model inference bookings now top $1.15 billion a year. When open models run faster and cheaper, more companies pick them, and more capital follows. That is what this round signals.

Does open-weight model funding affect my HR software?
Indirectly, yes. Most HR tools now run AI features on top of these models. As the cost of running them falls, the AI features in your HR stack should get cheaper or more capable. Ask your vendors which models they use, and whether your price reflects their falling costs.

What should HR leaders do before deploying AI agents?
Treat it as a governance question first. Before an agent touches payroll or employee records, ask how you would test, verify, and monitor it. Workday’s new Agent Passport checks agents against standards like the NIST AI Risk Management Framework. Use that idea even if you are not a Workday customer, and demand proof an agent is safe before you deploy it.

Not to be considered as tax, legal, financial or HR advice. Regulations change over time so please consult a lawyer, accountant  or Labour Law  expert for specific guidance.