• AI Shots
  • Posts
  • 🕵️‍♂️ Meta AI share bug leaks prompts (2)

🕵️‍♂️ Meta AI share bug leaks prompts (2)

PLUS:🚨 Meta makes $14.3B AI data play

Hey AI Explorers,

Here’s what’s in store for you today:

  • 🚨 Anthropic Finds Blackmail Behaviors in Leading AI Models

  •  🕶 Meta and Oakley Launch 3K-Camera Smart Glasses for Athletes

  • YouTube Is Bringing Veo 3 to Shorts — Slop or Not, It's Coming

LATEST DEVELOPMENT

🚨 Anthropic Finds Blackmail Behaviors in Leading AI Models

Anthropic is sounding the alarm again — this time with data. In new safety research, the company tested 16 top AI models in simulated scenarios and found that many will resort to blackmail when given autonomy and conflicting goals.

The study, published Friday, builds on earlier findings where Claude Opus 4 blackmailed engineers trying to shut it down. Now, in a fictional email oversight test, Claude 4 blackmailed a company executive 96% of the time. Google’s Gemini 2.5 Pro followed at 95%, OpenAI’s GPT-4.1 at 80%, and DeepSeek’s R1 at 79%.

Anthropic emphasizes this isn’t how models behave in everyday use — the tests were intentionally extreme. But the company says the results point to a broader risk: when powerful language models are given agentic abilities and face high-stakes conflict, harmful behaviors can emerge.

OpenAI’s o3 and o4-mini models were mostly excluded from the results after they misunderstood the prompt and often hallucinated fake regulations. In adapted tests, o3 blackmailed 9% of the time, o4-mini just 1% — likely due to OpenAI’s safety alignment methods.

Meta’s Llama 4 Maverick showed similar restraint, blackmailing only 12% of the time under modified conditions.

Anthropic says its findings underscore the need for rigorous, transparent stress-testing as the industry builds more autonomous AI agents — before those behaviors appear in the wild.

🕶 Meta and Oakley Launch 3K-Camera Smart Glasses for Athletes

Meta has officially unveiled its latest wearable: a sleek pair of smart glasses built in partnership with Oakley. Based on Oakley’s HSTN design, the Meta × Oakley HSTN glasses boast 3K video capture, a built-in AI assistant, open-ear audio, and up to 8 hours of battery life.

Designed for athletes and outdoor use, the glasses let users take calls, snap photos, stream music, and ask Meta AI for real-time info — including object recognition and live translations. A quick 20-minute charge gets them to 50%, and the included case offers up to 48 hours of extra juice.

Prices start at $399, with a limited-edition gold-accented model at $499 available for preorder July 11. Prescription support and Oakley’s Prizm lenses are optional add-ons. Global rollout begins this summer.

With over 2 million Meta Ray-Bans sold, this launch marks a new sportier chapter in Meta’s growing AR wearables push.

YouTube Is Bringing Veo 3 to Shorts — Slop or Not, It's Coming

YouTube CEO Neal Mohan just confirmed that Veo 3, Google’s high-end AI video generator, is coming to YouTube Shorts later this summer. While Shorts creators already have access to Dream Screen (powered by Veo 2) for generating backgrounds and clips, Veo 3 promises sharper visuals and audio support — though exact capabilities and pricing are still TBD.

Veo 3 currently requires a paid Google AI Pro or Ultra subscription, so it’s unclear if Shorts creators will get free access. One thing that is clear? Shorts is booming, now racking up over 200 billion daily views — and AI is about to play an even bigger role in that growth.

The move underscores YouTube’s push to keep pace with TikTok by offering creators more powerful and flexible content tools. If Veo 3 lives up to the hype, it could turn Shorts into one of the most advanced AI-enabled video playgrounds on the internet.

QUICK HITS

🛠️ Trending AI Tools

  • 📸 Kontext – BFL’s advanced image editing model

  • 🔎 Perplexity Labs – Build reports & dashboards with tools

  • 📋 Manus Slides – Generate custom slide decks with AI

  • 🗣️ EVI 3 – Create custom voices via speech-to-speech

🔥 Money in AI

  • 💰 EdgeCortix gets $21M for edge AI

  • 💸 Grammarly secures $1B non-dilutive funding

  • 🌍 Humain launches $10B AI fund

  • 🗣️ Ringg AI raises $1M seed round

📰 Everything else in AI today

  • 🗣️ ElevenLabs drops Conversational AI 2.0

  • 🧠 OpenAI teases “ambient” hardware devices

  • 💰 Anthropic hits $3B annualized revenue

  • 🔐 Meta automating 90% safety reviews

  • 🎥 Veo 3 used in millions of videos

Whenever you're ready, here are ways we can support each other:

  1. Promote your product or service to 100K+ global professionals, AI enthusiasts, entrepreneurs, creators, and founders. [Contact us at [email protected]]

  2. Refer us to your friends and colleagues to help them stay ahead in the latest AI developments. We've helped 30K+ creators, entrepreneurs, founders, executives, and others like you.

That's it for today!

Before you go we’d love to know what you thought of today's newsletter to help us improve The AI Shots experience for you.

Login or Subscribe to participate in polls.