Anthropic hits $65B run rate, DeepSeek bets on agents, EU AI Act enforcement begins
Anthropic's revenue run rate reportedly crossed $65 billion ahead of a likely IPO. Meanwhile DeepSeek launched V4 Pro as an open-source agent platform, and EU AI Act enforcement went live on 2 August.
Every line links to a primary source in the full briefing.
01Anthropic's revenue run rate reportedly tops $65 billion ahead of a possible IPO
Axios and Bloomberg reported that Anthropic's annualised revenue run rate has surpassed $65 billion. The figure suggests the company behind the Claude family of models has seen very rapid enterprise adoption. No IPO date has been confirmed, but the milestone is being watched as a pre-IPO signal.
When one of the two leading AI labs is growing at this pace, it accelerates product releases, pricing competition and enterprise feature development across the whole industry, all of which affects what tools and prices are available to smaller businesses.
Source: Axios
02DeepSeek V4 Pro launches as an open-source agent platform, not just a model
DeepSeek released V4 Pro with a plugin-based architecture and open-source tooling called Harness. The design signals that DeepSeek is positioning itself as a foundation for building AI agents, where third parties add capabilities as plugins, rather than competing purely on model quality. The release comes with significant upgrades to agent handling.
An open-source agent platform from DeepSeek gives smaller development teams a free, extensible base to build business automation on, which could reduce the cost of custom agent projects.
03EU AI Act enforcement started on 2 August, transparency rules now apply
The European Commission's AI Office began enforcing the AI Act on 2 August 2026, together with national authorities. A key new requirement is that certain AI systems must inform users when they are interacting with AI rather than a human. The Commission also published official guidelines to help providers and deployers meet the transparency obligations.
If your business deploys a chatbot, automated decision tool, or any AI that interacts with customers in the EU, you may now be legally required to disclose that it is AI, and non-compliance can trigger enforcement action.
04OpenAI says defenders have a current window of advantage over AI-assisted attackers
OpenAI published a report on AI and cybersecurity arguing that, right now, defenders who use AI tools can move faster than attackers who rely on the same technology. The report outlines how OpenAI is hardening its own systems and what practical steps security teams can take today to maintain that advantage.
For any business running its own IT or using cloud AI tools, the report is a concrete prompt to review whether your security team is already using AI-assisted defences before attackers close the gap.
Source: OpenAI
05Meta releases Muse Glimmer, an open-source multimodal agent that runs locally
Meta published Muse Glimmer on Hugging Face. It is described as a local, agentic, multimodal and open-source model, meaning it can process different types of input, take actions, and run on a user's own hardware without sending data to an external server. The release was posted roughly eight days ago but has only recently gained wider attention.
A capable multimodal agent that runs entirely on local hardware is relevant to businesses that handle sensitive data and cannot send it to third-party cloud services.
Source: Hugging Face
06SMBs are adopting AI fast but 57% say they cannot tell whether it is working
A report highlighted by Los Andes found that small and medium businesses are moving quickly to adopt AI tools, but 57 percent of them still do not know whether the investment is actually producing results. The gap between adoption and measurable impact appears to be a widespread problem across the sector.
If you have rolled out AI tools without setting clear metrics beforehand, you are likely in that 57 percent, and the first practical step is defining what success looks like before adding more tools.
Source: Los Andes
Sources
8 links- 01Anthropic's revenue run rate reportedly surpasses $65 billion pre-IPOAxios
- 02DeepSeek Harness Open Source: Everything Is a Plugin — the Bet Is an Agent Platform, Not a ProductPandaily
- 03DeepSeek V4 Pro Launches With Major Agent Upgrades and Open-Source HarnessMemeburn
- 04The Defender’s WindowOpenAI
- 05Las PyMEs se lanzan a la inteligencia artificial, pero el 57% todavía no sabe si le sirveLos Andes
- 06Commission starts enforcing AI Act rules and new transparency requirements on 2 AugustEuropean Commission
- 07Commission publishes guidelines on transparency obligations for providers and deployers of certain AI systemsEuropean Commission
- 08Meta is back with Muse Glimmer: local, agentic, multimodal, and open sourceHugging Face