Hallucination Station: Issue 12

🌟 Editor's Note

Congrats on joining the community! If you like what you see then please share with your friends so we can iterate and improve and bring you the best newsletter possible!

🚀 This Weeks Top Stories in AI

🧠 Pause on MLK-depictions by OpenAI

• OpenAI has temporarily suspended its AI systems from generating content depicting Martin Luther King Jr. at the request of the Estate of Martin Luther King Jr., Inc. (“King, Inc.”) due to concerns over disrespectful imagery.

• The move follows instances where users created unacceptable portrayals of Dr. King, prompting this preventive moderation step to uphold dignity and respect.

• While this restriction affects specific depiction types, OpenAI emphasizes its broader commitment to safety and respectful use of its generative AI tools across all scenarios.

👉 Read More About It

🎬 Actors vs. AI: OpenAI teams up with SAG‑AFTRA and Bryan Cranston to tighten likeness rules in Sora 2

• OpenAI has acknowledged that its video-generation tool Sora 2 allowed users to replicate Bryan Cranston’s voice and likeness without consent, prompting intervention by SAG-AFTRA.

• The company committed to an opt-in policy for voice and likeness usage, strengthened guardrails, and pledged to promptly address misuse complaints. Talent agencies, rights-holders and OpenAI co-signed a statement affirming that all performers have the right to determine how and whether they can be simulated, signalling broader industry shifts.

👉 Read More About It

🚀 Major Compute Leap: Anthropic & Google in Billion-Dollar TPU Deal

• Anthropic has secured access to up to 1 million Google TPU chips, worth tens of billions of dollars, for its AI infrastructure.

• The deal brings over 1 gigawatt of computing capacity online by 2026, significantly scaling Anthropic’s ability to train its large language models.

• For Google Cloud it strengthens its role as a strategic alternative to NVIDIA in the AI chip ecosystem, while Anthropic diversifies beyond a single-vendor model.

👉 Read full breakdown

🌐 Something to browse: ChatGPT Atlas lands

• Built by OpenAI, ChatGPT Atlas is a new web browser for macOS that embeds ChatGPT directly into the browsing experience — letting you ask, analyze or edit right within the page.

• Key features include a sidebar for context-aware help, “Browser memories” to remember visited sites (optional and user-controlled), and an “Agent mode” (for paid tiers) where ChatGPT can navigate and act for you.

• While initial launch is macOS-only (with Windows, iOS/Android versions coming), the move signals OpenAI challenging traditional browser dominance and shifting AI from “assistant app” to “integral browser layer”.

👉 Source: OpenAI

🦄 Other Bits

• AWS Postmortem: A single DNS race condition brought Amazon’s cloud empire to its knees

• Meta cuts 600 jobs amidst AI expansion.

• Boom Boom Boom, now let me hear you say (AI) Bubble!

🔥Meme-while in AI

(Coming next week - send me your favorite AI memes)

🏆Tool of the Week (not sponsored)

My thoughts on voice as an interface ebb and flow. But Willow makes a pretty compelling argument. On the one-hand, I struggle to see voice taking over completely - typing is a great way to preserve privacy in public spaces. I’m not going to break up with my partner over Voice text while on a public bus! I don’t want the strangers sat next to me to know how despicable I am. But for the shame-free content I want to release from my brain, voice is hella efficient. Check out Willow.

Did You Know? The first computer bug was literally a bug—in 1947, Grace Hopper found a moth trapped in a Harvard Mark II computer, coining the term "debugging" in the process.

Till next time,