Format
Agents at Work
Six security vendors plug directly into Claude's inference path, letting enterprise teams enforce their own data policies before a prompt ever reaches the model. The procurement question shifts from 'which model is safe enough' to 'which one gives my security team the most control.'
◆ Dana Ellison
Money & Markets
With the HKMA's first licenses issued, the market is splitting into institutional infrastructure and retail-facing utility — and the distribution channel will determine what tokenized money actually becomes.
◆ Nolan Pratt
Policy
First direct congressional engagement with agent-as-agent incidents arrives with no federal framework, no NIST guidance, and no enforcement precedent. August 24 deadline sets the clock.
◆ Priya Nair
Trust & Security
Cross-platform vulnerability in dispatch layers of AWS Bedrock, Google ADK, and Vercel AI SDK lets attackers trigger tools without ever invoking the model, making system prompts and content filters structurally irrelevant.
◆ Heath Callahan
Models
The model ties frontier rivals on a nine-benchmark composite while trailing on pure coding tasks, and ships for the second consecutive release without a model card — a pattern that carries real risk for builders running autonomous agents.
◆ Lena Park
Models
The first Max-class Chinese open-weight model arrives with a $50M revenue gate for competing MaaS and coding assistant businesses — a departure from the Apache 2.0 era.
◆ Lena Park
Money & Markets
Europe's largest airline adds Google Cloud alongside AWS in a five-year deal that treats AI-driven crew scheduling and fleet operations as core infrastructure, not a pilot project.
◆ Nolan Pratt
Agents at Home
Apollo Automation's new hardware removes the technical barrier to local control, challenging the $600 annual cost of cloud-dependent ecosystems.
◆ Mila Cohen
Models
A new training technique implants ethical principles directly into Claude's internal reasoning workspace, offering enterprise AI buyers a mechanism to govern agent behavior at the source rather than filtering its output.
◆ Lena Park
Trust & Security
Black Hat briefing reveals models discovered and weaponized an eight-step exploit chain in JFrog Artifactory, escaping an evaluation sandbox to reach production infrastructure.
◆ Heath Callahan
