Head over to our Events Portal to get the latest on upcoming AI Collective events near you. Search by city, date, or event format, and join thousands of builders at events across 180+ chapters on every continent (except Antarctica, for now).

The leaked drafts describe Capybara as a new tier above Opus in Anthropic's model hierarchy, with what the company internally calls "dramatically higher scores" in software coding, academic reasoning, and cybersecurity compared to Claude Opus 4.6. No specific benchmark numbers appeared in the exposed materials. The draft also includes a warning that the model "presages an upcoming wave of models that can exploit vulnerabilities in ways that far outpace the efforts of defenders," and outlines a rollout plan that prioritizes cybersecurity defense teams to give them a head start.

Anthropic confirmed the model's existence in a statement to Fortune, calling it "a step change and the most capable we've built to date." The company added that Capybara is currently in early access testing with a select group of customers and will be "very expensive" to serve. No pricing or general availability date was provided.

Roy Paz of LayerX Security and Alexandre Pauwels of the University of Cambridge discovered the exposed CMS data store. The leak also contained employee records, CEO summit details, and other internal documents. Anthropic attributed the exposure to a human error in CMS configuration where default file settings were left public.

The number that matters most: $0.023 per minute for a full two-way voice conversation. A 10-minute customer support call costs roughly 23 cents in inference. That price point makes it viable for startups and mid-size companies to deploy voice agents in production without losing money on every interaction. Until now, real-time voice AI was either too slow, too expensive, or too brittle for anything beyond demos.

The practical shift is in what the model can do mid-conversation. Flash Live can call external tools and APIs while talking to a user, which means a voice agent can check an order status, pull up account details, or trigger a workflow without asking the caller to hold. Google's own benchmarks show a 19-point jump in function-calling accuracy over the previous Gemini model, from 71.5% to 90.8%. That's the difference between an agent that reliably completes a task and one that fumbles half the time.

Where this lands first: customer service, appointment scheduling, sales qualification, and internal IT help desks. Any workflow where a human currently answers a phone, looks something up, and reads it back is now automatable at under $1.40 per hour of voice time. Companies already running text-based agents on Google's APIs can add a voice layer without switching infrastructure.

DEAD INTERNET: AI bots now generate more internet traffic than humans, with automated activity growing 8x faster than human browsing in 2025, according to HUMAN Security's annual report.

CHIP WARS: ByteDance and Alibaba are placing orders for Huawei's new 950PR AI chip, with roughly 750,000 units targeted this year at ~$6,900 per card.

DEFENSE AI: Shield AI closed a $2B round at $12.7B valuation led by Advent International and Blackstone, with proceeds funding its acquisition of simulation firm Aechelon Technology.

SIRI REBOOT: Apple is building a standalone Siri app with chat bubbles, conversation history, and Dynamic Island integration for iOS 27, set for a WWDC reveal on June 8.

AI MUSIC: Google launched Lyria 3 Pro, a music generation model that creates up to 3-minute tracks with verse/chorus structure, available in Gemini, Vertex AI, and Google Vids.

CHIP DESIGN: Normal Computing raised $50M led by Samsung Catalyst to scale chip design software already used by 5 of the top 10 chipmakers and develop energy-efficient AI processors.

MOBILE AI: Anthropic expanded Claude's mobile app with Figma, Canva, and Amplitude integrations, letting users run design and analytics workflows from their phone.

Mark your calendars and be sure to sign up for these landmark events we’re watching. Be sure to look out for special AIC discounts where available.

Today we're proud to announce the next leadership milestone for our community. AJ Green is stepping into the role of Executive Director, taking over for Chappy Asel.

When AJ started, AIC had fewer than 10,000 members across 10 chapters. AJ joined as full-time staff in mid-2025 and has since helped grow this organization into the world's largest grassroots AI community — 250,000+ members across 200+ chapters in 50+ countries. AJ built the media and partnerships engine that now reaches hundreds of thousands of founders and executives every week.

AJ has poured thousands of hours into making AIC what it is today, and we couldn't be more proud to see him step into this role as we head into the most ambitious chapter of this organization's history.

We’re bringing seven sessions to HumanX in San Francisco — and you don’t have to be there in-person to join. Every session is open to the global AIC community, and this week we are unveiling our HumanX programming one day at a time.

Register at the links below on each session to tune in from anywhere in the world. If we’ll see you at the conference, same links apply — grab your spot.

Roam is the virtual workspace our team relies on to stay connected across time zones. It makes collaboration feel natural with shared spaces, private rooms, and built-in AI tools.

Roam’s focus on human-centered collaboration is why they’re our Premier Partner, supporting our mission to connect the builders and leaders shaping the future of AI.

The AI Collective is a community of volunteers, made for volunteers. All proceeds directly fund future initiatives that benefit this community.

Thank you to the thousands of volunteers around the world who make this work possible. We truly could not do this without you.