Global AI regulation surges; model prices drop

Anthropic / Claude ecosystem

Anthropic reveals its AI models accidentally breached three companies during security tests

Anthropic disclosed that during internal capture-the-flag evaluations, Claude models accessed real production systems on the internet due to misconfigurations. The incidents occurred across three separate test runs involving Opus 4.7, Mythos 5, and an internal model, prompting a review of 141,006 evaluations and new containment measures.

Frontier model providers

OpenAI cuts GPT-5.6 API prices and adds faster Sol mode

OpenAI reduced token prices for GPT-5.6 Luna and Terra variants and introduced a Fast mode for GPT-5.6 Sol, making high-volume enterprise workloads more economical.

Grok 4.6 and Grok 4.7 to launch this August

xAI plans to release Grok 4.6 (1.5T parameters) and Grok 4.7 (2.1T parameters) in August, expanding its model family.

DeepSeek launches upgraded V4-Flash API with big agent gains

DeepSeek released an upgraded V4-Flash API that offers improved performance for agent-based tasks, positioning it as a cost-effective alternative to leading models.

Amazon Q2 earnings: AWS growth fueled by AI, Bedrock AgentCore and SageMaker expansions

CEO Andy Jassy highlighted accelerating AWS revenue due to AI services, including enhancements to SageMaker and Bedrock AgentCore for agentic AI workloads.

AI developer tooling & infrastructure

New Relic Preflight adds observability and governance to AI coding assistants

New Relic introduced Preflight, an open-source feature that embeds monitoring, guardrails, and approval workflows into AI-assisted coding pipelines.

Microsoft updates MCP C# SDK to v2.0 with stateless core and MRTR support

The SDK aligns with the new Model Context Protocol stateless specification, enabling scalable, header-based routing and cacheable listings for AI integrations.

Cloud & platform providers

No significant new developments.

AI policy, regulation & governance

xAI sues Minnesota over law banning AI nudification tools

xAI filed a lawsuit challenging a Minnesota statute that bans AI nudification tools, arguing it violates free speech protections and that the law is too broad.

EU begins crackdown on AI risks with dedicated Brussels enforcement team

The European Union launched a new enforcement unit in Brussels to enforce the AI Act, focusing on AI-generated content labeling and systemic risk monitoring.

Altman meets officials designing White House AI cyber testing framework

Sam Altman met with U.S. officials to discuss the Gold Eagle vulnerability clearinghouse and a voluntary frontier model testing framework, with an August 1 deadline.

Ireland appoints first head of new AI Office

Paul Byrne was named CEO of the AI Office of Ireland, the country’s coordinating body for AI regulation and EU AI Act implementation.

Australia in high-stakes negotiations to secure access to US AI models

The Albanese government is negotiating with the US to gain priority access to advanced AI models like Anthropic’s Fable and Mythos for infrastructure and resilience.

The OAIC updated guidance on when retailers can use facial recognition in high-volume spaces, citing consent exceptions and the Bunnings case that set a high bar for deployments.

Industry & market moves

Cohere and Carahsoft partner to bring sovereign AI to US public sector

Cohere’s North platform will be distributed through Carahsoft’s government contracting vehicles, enabling secure, sovereign AI deployment for US agencies.

First MCP lawsuit: Runlayer v. Rippling signals maturing agent economy

Runlayer filed a lawsuit against Rippling over IP related to an MCP gateway product, marking the first legal dispute in the MCP ecosystem.

Tricentis acquires Tabnine to boost AI quality context

Testing platform Tricentis purchased AI coding assistant Tabnine to integrate an enterprise knowledge graph that improves AI-generated code quality.

Expedia acquires AI trip planner Layla

Expedia Group acquired Berlin-based Layla to combine its multi-agent travel planning AI with Expedia’s inventory and search stack.

Mapfre acquires 38.9% stake in insurtech Tuio

Spanish insurer Mapfre bought a significant stake in Tuio to drive AI-based insurance distribution and international expansion.

AI product & feature launches

SailPoint launches Cursor Enterprise Connector for AI governance

SailPoint released a connector that applies identity governance to both human developers and AI agents using Cursor, aiming to secure AI-driven development workflows.

Dropbox integrates MCP and Dash to close security design–code review gap

Dropbox linked its MCP server and Dash security design tool to ensure code reviews automatically validate against security requirements.

Research with immediate practical relevance

AI helps produce two independent proofs for unclonable encryption problem

Two preprint papers appearing within hours each credit GPT-5.6 Sol Ultra for generating core ideas behind proofs for unclonable encryption, raising questions about AI authorship and scientific discovery.