AI Policy Heats Up

By the Orchestra editors · 4 min read

TL;DR

Albanese flags major AI policy push as early as next month, with copyright workaround for local model training on the table
Anthropic accuses Alibaba of massive Claude distillation attack using 25,000 accounts
OpenAI reveals custom inference chip Jalapeño built with Broadcom
SMS sender ID deadline July 1: unregistered business texts will show as 'Unverified'

AI · Worldwide

Anthropic accuses Alibaba of massive Claude cloning attack

Anthropic says Alibaba used roughly 25,000 accounts to extract Claude's capabilities across 28.8 million exchanges in what the company calls the largest distillation attack it has faced. Anthropic is pushing for penalties.

OpenAI unveils Jalapeño, its first custom inference chip

Built in partnership with Broadcom, the ASIC is designed specifically for LLM inference at scale. It arrives as OpenAI tries to reduce reliance on Nvidia and manage surging demand across ChatGPT and Codex.

Claude gaining ground on ChatGPT among paying users

Data shows consumers willing to pay for AI assistants are increasingly choosing Anthropic's Claude, even as ChatGPT maintains its overall market lead. Worth watching for teams evaluating which model to standardise on.