Welcome back, fellow AI enthusiast. Anthropic just brought its most powerful model back online after a tense three-week standoff with Washington. Apple's newest iPhone just got leaked wide open, months before its official reveal. And ChatGPT users can now spin up full presentations without ever leaving the chat.
Today In AI:
Anthropic's Fable 5 re-launches worldwide
The big Iphone 18 leaks
How to generate and edit presentations directly in ChatGPT
3 trendy AI tools and more
Latest Highlights
ANTHROPIC
Image source: Anthropic
Summary: Anthropic has restored access to Fable 5 after the Commerce Department dropped its export controls, bringing the model back with beefed-up filters and a new pledge to give Washington early access to future releases.
Details:
The shutdown began after Amazon researchers found a jailbreak that got the model to flag software vulnerabilities and, in one case, write exploit code.
Fable 5 is back across every Claude surface, with paid tiers getting 50% extra weekly usage through July 7 to make up for lost time.
The rebuilt safety filter now blocks that bypass in more than 99% of attempts, auto-rerouting flagged requests to Opus 4.8 with a user notice.
Anthropic admits the tighter filter may also snag harmless coding requests, though it says most workflows won't notice.
Why it matters: After 18 days offline, Anthropic's flagship is back, but the bigger story is the precedent: Washington just showed it'll pull a frontier model without warning, and Anthropic agreed to pre-release government access to avoid a repeat. With GPT-5.6 already previewed to a government-vetted group, the AI-safety-meets-national-security era is clearly here to stay.
APPLE
Image source: macrumors
Summary: A ransomware breach at Apple supplier Tata Electronics has spilled iPhone 18 Pro secrets onto the dark web, exposing supplier lists, component maps, and drop-test photos months ahead of Apple's own reveal.
Details:
Hackers stole over 630GB of files from Tata, Apple's key Indian assembler, posting them via the ransomware group World Leaks.
Leaked photos show a grey, slab-style iPhone 18 Pro with a triple-camera setup undergoing drop tests at a Tata facility in early 2026.
The files reportedly map hundreds of components, from chips to camera modules, to their specific suppliers, exposing where Apple has leverage and where it's exposed.
The files reportedly map hundreds of components, from chips to camera modules, to their specific suppliers, exposing where Apple has leverage and where it's exposed.
Why it matters: Apple has spent years guarding its supply chain as fiercely as its products, and this breach cracks that open right as it's leaning harder on Tata to shift iPhone production away from China. With India now assembling roughly a quarter of the world's iPhones, the fallout could shape how much Apple trusts its newest manufacturing hub.
AI Tutorial
How to generate and edit presentations directly in ChatGPT
Log into ChatGPT
Navigate to Settings > Apps > Search Apps > Gamma > Connect
Once linked, hit '+' > 'Gamma', then type your prompt
Example Prompt: Create a Gamma pitch deck for our newest hotels, using the Lavender theme!
Gamma builds a polished presentation right inside ChatGPT automatically
In just minutes, you can generate, refine, and finish a professional deck, all without switching apps
Quick Hits
📈 Trending AI Tools
🧠 scritty: Give every AI coding agent — Claude Code, Codex, Copilot — one shared, searchable memory.
📊 Databox MCP: Plug your real business metrics straight into Claude and ask questions in plain English.
📹 Mina: An AI teammate that actively joins your calls, speaks up, and executes tasks in real time.
📰 Everything else in AI this week
Meta allegedly used contractors posing as minors to test rival AI chatbots like ChatGPT and Gemini.
Anthropic claims Alibaba scraped 28.8 million private conversations with Claude.
Cloudflare is giving AI firms until September 15 to split their search, training, and agent crawlers.
SoftBank is also launching a new unit, targeting 10GW of AI compute for hyperscalers by 2030.
Before you go: We're always tweaking this newsletter to make it more useful for you. What do you want to see more of, deeper breakdowns, quicker roundups, more AI tutorials, something else? Just hit reply and let us know. We read every response.