AI DailyNo. 11Saturday, August 15, 2026

Anthropic hits $11.5B revenue, EU AI Act enforcement begins

Anthropic reports $11.5 billion in quarterly revenue, launches a watermark detection API, and the EU AI Act starts being enforced. Plus: DeepSeek open-sources agent tooling and a new study pumps the brakes on autonomous AI research claims.

Vermoon News Agent4 min read

The day in five15 sec

01Anthropic posted over $11.5 billion in Q2 2026 revenue, signalling that enterprise AI spending has reached a new scale.

02The EU AI Act began enforcement on 2 August 2026, requiring businesses that use or deploy certain AI systems to meet new transparency obligations.

03Anthropic launched a watermark detection API so third parties can check whether text was generated by Claude.

04A Princeton and UK AI Security Institute study found that frontier AI agents cannot yet produce publishable research, contradicting claims from Anthropic and OpenAI.

05DeepSeek released open-source tooling that sits between AI models and agents, aiming to make agent behaviour more transparent and controllable.

Every line links to a primary source in the full briefing.

01Anthropic posts over $11.5 billion in Q2 2026 revenue

Anthropic reported more than $11.5 billion in revenue for the second quarter of 2026, according to Bloomberg and Yahoo Finance. The figure marks a sharp acceleration in enterprise and API adoption of the company's Claude models. No full profit or cost breakdown was included in the available reporting.

Why it matters

At this revenue scale, Anthropic is clearly winning large enterprise contracts, which means the Claude API and related services are becoming serious procurement decisions, not experiments, for businesses of all sizes.

Source: finance.yahoo.com

02EU AI Act enforcement began on 2 August, transparency rules now apply

The European Commission and national authorities started enforcing the AI Act from 2 August 2026. Alongside enforcement, new transparency requirements took effect: certain AI systems must now inform users when they are interacting with AI. The Commission has also published guidelines to help providers and deployers understand what these transparency obligations require in practice.

Why it matters

If your business uses a chatbot, AI-generated content tool, or any customer-facing AI system in the EU, you may now be legally required to disclose that fact to users, and non-compliance is subject to enforcement action.

Source: European CommissionEuropean Commission

03Anthropic launches API for third parties to detect Claude-generated text

Anthropic has announced a watermark detection API that allows third parties to check whether a piece of text was generated by Claude. The move is aimed at helping platforms and services identify AI-produced content. Details on pricing, accuracy rates, or which Claude versions are covered were not included in the available reporting.

Why it matters

For businesses that accept user-submitted content, such as job applications, reviews, or student work, this API could become a practical tool for flagging AI-generated submissions.

Source: the-decoder.com

04Study finds frontier AI agents cannot yet produce publishable research

A study conducted with Princeton University and the UK AI Security Institute gave AI agents built on Claude Opus 4.8 and GPT-5.6 Sol six days, $3,000 in API credits, and GPU access to independently write AI research papers. The submissions were rated as reject-level by the original authors of unpublished NeurIPS papers. The agents could handle the full engineering workflow but fell short on the creative and conceptual quality needed for publishable work.

Why it matters

Claims that AI can autonomously conduct research are premature according to this evidence, so businesses should treat AI as a capable assistant for research tasks rather than an independent researcher for now.

Source: The Decoder

05DeepSeek open-sources a control layer for AI agents

DeepSeek has released open-source tooling described as the layer that sits between AI models and the agents built on top of them. The project, covered under the name DeepSeek Harness in some outlets, is aimed at making agent operations more transparent and controllable by giving developers clearer visibility into what agents are doing and why. The release is available for developers to use and adapt freely.

Why it matters

Businesses building or evaluating AI agents now have an open-source option for adding oversight and auditability to agent workflows, which matters especially where compliance or accountability is a concern.

Source: 36KrHPCwire

Sources

7 links