Sam Altman on Erotica: “OpenAI is not the moral police of the world.”

Welcome back! The boundaries of AI are getting messier. OpenAI is rewriting the rules of what adults can do with chatbots, scientists are trying to simulate life itself, and Silicon Valley’s moral debate has turned into open warfare. Somewhere in that noise sits the question that defines this era: who decides how far intelligence should go?

In today’s Generative AI Newsletter:
• Sam Altman says OpenAI isn’t the “moral police of the world”
• Scientists attempt to build life inside a computer
• Silicon Valley turns on the AI safety movement
• Facebook’s AI learns to edit photos you haven’t shared yet

OpenAI CEO Sam Altman has defended plans for an “Adult Mode” in ChatGPT after a social media storm over the company’s stance on erotica. Altman clarified that the change is meant to expand user freedom for adults while keeping safeguards for minors. He said the erotica example was “just one illustration” of allowing people more choice in how they use AI.

What Altman Said
• User freedom first. The new mode is meant to let verified adults explore a wider range of content and expression.
• Safety for minors stays firm. Altman said the company will continue prioritizing protection for teenagers and those dealing with mental health challenges.
• Boundaries remain. Content that causes harm or targets others will still be prohibited.
• OpenAI’s stance. “We are not the elected moral police of the world,” Altman wrote, adding that the company aims to set limits similar to how society defines R-rated boundaries.

The debate highlights how AI is starting to test the edges of personal expression. What once sounded like a technical discussion about moderation is becoming a conversation about autonomy and culture. OpenAI seems intent on walking that narrow path where freedom, taste, and responsibility all collide.

Special highlight from our network

This milestone reflects their commitment to helping enterprises deploy AI responsibly with trust, governance, and security at the core.

What Tumeryk delivers:

👉 Experience what trusted AI looks like

Book a Discovery Call

Researchers are using AI to build a complete digital model of a living human cell. The project aims to recreate how real cells function so scientists can test drugs, study diseases, and simulate biology entirely inside a computer. What once seemed unrealistic is now becoming technically possible because of advances in computation and data modeling.

What’s Happening
• AI replaces trial and error. Teams at Google DeepMind and the Chan Zuckerberg Initiative are training models that can predict how cells react to drugs in seconds, reducing years of lab work to digital experiments.
• Learning from data. Instead of building equations by hand, researchers feed vast cellular datasets into AIs that learn to forecast how real cells behave.
• Cross-species insight. One model trained on twelve species accurately predicted how cells from unfamiliar species would respond.
• A new foundation for biology. The vision is to make cell biology mostly computational, where experiments begin in simulation before moving to the lab.

Scientists now talk about the virtual cell with the same excitement once reserved for the Human Genome Project. “You build models that learn directly from data,” said Stanford’s Stephen Quake, whose team helped define the field. “The dream is to turn cell biology from ninety percent experimental to ninety percent computational.”

The fight over AI safety has turned personal. Tech leaders including White House AI and Crypto Czar David Sacks and OpenAI Chief Strategy Officer Jason Kwon have accused AI safety groups of acting out of self-interest rather than public concern. Their comments set off a storm online and have left several nonprofit leaders uneasy, some asking to stay anonymous out of fear of retaliation.

What Sparked It
• Sacks vs. Anthropic. Sacks accused Anthropic of “regulatory capture,” claiming the company supports laws like California’s SB 53 only to block smaller rivals. Anthropic was the only major lab to endorse the bill, which requires safety reporting from large AI firms.
• OpenAI’s legal push. Kwon confirmed that OpenAI sent subpoenas to multiple AI safety nonprofits, including Encode, seeking their communications with Elon Musk and Mark Zuckerberg. He said the goal was to uncover potential coordination after Musk sued OpenAI over its nonprofit status.
• Internal dissent. OpenAI’s head of mission alignment, Joshua Achiam, criticized the subpoenas, writing, “At what is possibly a risk to my whole career I will say: this doesn’t seem great.”
• The larger divide. Many in the AI safety community say these tactics are meant to intimidate. “OpenAI is trying to silence critics,” said Brendan Steinhauser of the Alliance for Secure AI.

On one side are those chasing scale and profit. On the other are those trying to slow the machine long enough to understand it. For an industry that prides itself on innovation, it’s revealing how defensive it becomes when the conversation shifts from progress to restraint. The more Silicon Valley tries to spook its critics, the more it proves why the watchdogs are needed in the first place.

Facebook has begun rolling out a feature that lets Meta AI suggest edits to the photos still sitting in your camera roll. Users in the U.S. and Canada can now opt in to let the app access unposted images, which are then uploaded to Meta’s cloud for processing. The AI can create collages, themed edits, and personalized recaps that prompt users to share them on their feeds and stories.

How It Works
• Cloud access on request. Users see a permission prompt asking to “allow cloud processing,” enabling Meta’s AI to analyze and edit local photos.
• AI-generated creativity. The system offers ideas like collages, birthday layouts, or “AI restyling,” turning private images into share-ready content.
• Limited training use. Meta says the photos won’t be used for ads or AI training unless users choose to edit or share them publicly.
• The trade-off. By granting access, users still allow Meta to analyze image contents, facial features, dates, and objects, giving the company biological and behavioral data.

Meta frames the update as a creative feature, but the implications reach deeper. Allowing AI into the camera roll invites the platform into one of the few private spaces left on a phone. It’s a gentle reminder that in Meta’s world, even unposted moments are potential data points waiting to be seen.