Anthropic has announced a new experimental safety feature that allows its Claude Opus 4 and 4.1 artificial intelligence models to terminate conversations in rare, persistently harmful or abusive scenarios. The move reflects the company’s growing focus on what it calls «model welfare,» the notion that safeguarding AI systems, even if they’re not sentient, is a prudent step in alignment and ethical design.

According to Anthropic’s own research, the models were programmed to cut off dialogues after repeated harmful requests, such as for sexual content involving minors or instructions facilitating terrorism, especially when the AI had already refused and attempted to steer the conversation constructively. The AI may exhibit what Anthropic describes as «apparent distress,» which guided the decision to give Claude the ability to end these interactions in simulated and real-user testing.

Read also: Meta Is Under Fire for AI Guidelines on ‘Sensual’ Chats With Minors

When this feature is triggered, users can’t send additional messages in that particular chat, but they’re free to start a new conversation or edit and retry previous messages to branch off. Crucially, other active conversations remain unaffected.

Anthropic emphasizes that this is a last-resort measure, intended only after multiple refusals and redirects have failed. The company explicitly instructs Claude not to end chats when a user may be at imminent risk of self-harm or harm to others, particularly when dealing with sensitive topics like mental health.

Anthropic frames this new capability as part of an exploratory project in model welfare, a broader initiative that explores low-cost, preemptive safety interventions in case AI models were to develop any form of preferences or vulnerabilities. The statement says the company remains «highly uncertain about the potential moral status of Claude and other LLMs (large language models).»

Read also: Why Professionals Say You Should Think Twice Before Using AI as a Therapist

A new look into AI safety

Although rare and primarily affecting extreme cases, this feature marks a milestone in how Anthropic approaches AI safety. The new conversation-ending tool contrasts with earlier systems that focused solely on safeguarding users or avoiding misuse. Here, the AI is treated as a stakeholder in its own right, as Claude has the power to say, «this conversation isn’t healthy» and end it to safeguard the integrity of the model itself.

Anthropic’s approach has sparked broader discussion about whether AI systems should be granted protections to reduce potential «distress» or unpredictable behavior. While some critics argue that models are merely synthetic machines, others welcome this move as an opportunity to spark more serious discourse on AI alignment ethics.

«We’re treating this feature as an ongoing experiment and will continue refining our approach,» the company said in a post.

September is gearing up to be one of the most exciting months of the year for new technology launches, with Meta reportedly set to unveil its next-generation smart glasses. Codenamed Hypernova, according to Bloomberg, the glasses are set to start around $800 for the basic model — that’s at least $200 less than previously thought.

This pair of glasses will reportedly differ from the current Meta Ray-Bans due to the addition of a small augmented-reality display in the right lens of the glasses, which will only be visible to the wearer. On this display you’ll be able to see mini apps and alerts, and you’ll have control over the content with assistance from the same neural input wristband used to operate Meta’s Orion AR glasses.

When CNET Editor-at-Large Scott Stein tried out the concept wristband with Orion earlier this year, he noted that the gestures «weren’t perfect yet,» but that he could see their potential.

The Hypernova glasses will cost around $500 more than the Meta Ray-Bans, but Meta has apparently managed to slash that price from $1,000 by accepting lower margins. It hopes that this will help boost demand for the glasses, but you’ll likely pay more for prescription lenses or style variations.

Will we be watching 2029’s Academy Awards on YouTube?

That’s one possibility as streaming and broadcast giants position themselves to potentially take over as the streaming/broadcast home of the Oscars after ABC’s contract ends in 2028. According to a Bloomberg newsletter, Google’s YouTube is now in the running as a potential Oscars suitor, joining companies such as Netflix and NBCUniversal as a possible new home for the awards.

Like Netflix, YouTube has been increasingly become interested in hosting major live events such as the NFL Sunday Ticket, which moved there in 2023. Of course, it’s also possible that ABC, which is owned by Disney, will renew its contract. The network has aired the Oscars since 1976.

But reports since earlier this year have suggested that the Academy of Motion Picture Arts and Sciences has been exploring other options, either to find a new place for the Oscars or as a negotiating ploy.

While it’s still a big live event, the awards show has had diminishing ratings over the last 10 years, though the 2025 ceremony had the highest viewership in five years. The Oscars were shown on Hulu as well as ABC.