Tag: openai

OpenAI introduces more granular control of their personalities

New personality settings offer detailed controls for ChatGPT's personality.
New settings in ChatGPT lets you tweak the bot’s personality even more. (Picture: Screenshot)
Since GPT-5 it’s been possible to get ChatGPT to take on a lot of different personalities in how it interacts with the user.

Now, the controls for this has become much more detailed, with added «characteristics.»

You can control to what extent you want your chatbot to be Warm or Enthusiastic, and how much it uses Emoji and Headers & Lists, in addition to using a different «main voice and tone.»

The new settings are available to all users in the «Personalization» setting.

This is in line with OpenAI’s policy of making ChatGPT more personalized and customized for the end user — and we can probably expect more of this in the future.

Read more: writeup at The Verge.

OpenAI launches GPT Image 1.5, up to 4x faster and more precise, lifelike

This lifelike image of Sam Altman as Hugh Hefner was generated in one shot.
One-shotted: “Generate a picture of Sam Altman in the style of Hugh Hefner, in a smoking jacket with a cigar and a whiskey.” (Picture: GPT Image 1.5)
The new image model is faster by a wide margin, and way better at composing detailed, lifelike images.

It also «adheres to your intent more reliably» and keeps elements of the picture, like lighting, composition, and appearance across the editing phase.

The outputs are also much more lifelike, making it ever more difficult to discern AI images from reality.

Does it beat Nano Banana? Yes, it does, according to the almost gold-standard LMArena benchmark, where it instantly climbed to the top spot on «Text-to-image» and «Image Edit.»

At the same time, OpenAI is launching a new feature called ChatGPT Image on the web, which is a handy way to access predefined styles for your pictures and share those you like.

They are even seasonally aware enough to let you turn pictures into Christmas cards.

Read more: OpenAI’s announcement (includes lots of examples), ChatGPT Image. Writeups on Engadget, TechCrunch. Discussion on r/Singularity.

ChatGPT’s «adult mode» scheduled to debut in early 2026

"adult mode," or "erotica" as it was previously called, is coming soon to a ChatGPT near you.
«Let adults be adults,» said CEO Sam Altman in October 2025. (Picture: generated)
Altman had previously teased «erotica» for December this year, promising to treat «adults like adults.»

OpenAI now uses the term «adult mode» for the feature, and expects it to arrive in Q1 next year — once they are comfortable with their age-gating.

During the introduction of GPT-5.2, Fidji Simo, OpenAI’s CEO of Applications, said that they want to stick the age prediction on their models before debuting NSFW content, according to The Verge.

The idea behind age verification is primarily to impose stricter rules for teens on ChatGPT, but the practical upside is that they can also have looser rules for adults.

The tricky part is to not «misidentify adults,» Simo said.

Read more: Writeups on The Verge, and Gizmodo.

Disney invests $1 billion in cross-licensing deal for characters in Sora

Disney licenses beloved characters for use on Sora, and will get to use OpenAI tools in production.
While Disney is licensing to OpenAI, they are fighting other hard over others’ copyright violations. (Picture: andy orin, CC BY 2.0)
Users will be able to use Disney characters without breaking copyrights, in return for Disney getting to use the technology internally for three years.

The deal will license some 200 Disney-owned characters from the likes of Disney, Marvel, Pixar and Star Wars for use in Sora, free of charge (as in money) and charges (as in lawsuits).

The deal will also see Disney adopt ChatGPT and Sora internally, which is the other way of the license — helping Disney do everything from light sketches to development planning.

Continue reading “Disney invests $1 billion in cross-licensing deal for characters in Sora”

OpenAI launches GPT-5.2, and it beats Gemini 3 in a lot of benchmarks

GPT-5.2 smashes the ARC-AGI-2 leaderboard handily. (Picture: Screenshot)
As expected, OpenAI released its latest GPT this week, and this time, they are posting benchmarks.

It comes after OpenAI felt overtaken in the AI race by Anthropic and Google’s Gemini, when CEO Altman declared «Code Red» to overtake their rivals. That was a little over a week ago, and now the results are in.

Continue reading “OpenAI launches GPT-5.2, and it beats Gemini 3 in a lot of benchmarks”

OpenAI’s new enterprise report shows more adoption — and higher use

Power users of AI report more than ten hours of saved time per week.
Professionals are adopting AI at a fast clip, and getting more efficient, OpenAI finds. (Picture: Adobe)
The company evaluated real-world use from their Enterprise accounts and a survey of some 9,000 workers across 100 enterprises to come to this conclusion: AI at work seems to be rapidly taking off:

«The history of general purpose technologies» shows that the real value starts flowing once firms start adopting their capabilities at scale, explains OpenAI, and — Enterprise AI now appears to be entering this phase.

Continue reading “OpenAI’s new enterprise report shows more adoption — and higher use”

Head of ChatGPT says not doing ad tests, despite a Target notice this week

These were not ads, OpenAI says, and anyway, they have been shut off.
Many users were confronted with this graphic at the end of ChatGPT’s responses this week. OpenAI says they are now stopped, and that they weren’t ads. (Picture: screenshot)
People have been complaining about ads being sneakily inserted in their ChatGPT responses late this week, with a «Shop for home…» Target link showing up in their responses.

This comes hot on the heels of an engineer discovering ad code in the Android app — which left a lot of people cautious of such inserts.

It happened to enough people to notice, along with this writer, but it has since been turned off, says Mark Chen, Chief Research Officer at OpenAI:

Continue reading “Head of ChatGPT says not doing ad tests, despite a Target notice this week”

OpenAI loses privacy fight for ChatGPT message logs in NYT lawsuit

OpenAI is on the verge of losing its fight to keep their users' chat logs private, instead of turning them over to the NYT.
Your logs belongs belong to us, the NYT lawyers say, and the court agrees. (Picture: Adobe)
OpenAI has been fighting tooth and nail to preserve the privacy of their users’ messages in the lawsuit brought by The New York Times in 2023.

A new judgement could now mean they have to turn over more than 20 million chat logs, and many more messages, from the chatbot, reports Reuters.

The logs themselves should be anonymized by OpenAI in a way that pleases the court, but their content could be easy to pin down, and OpenAI has promised to appeal to the presiding Judge.

This is merely the discovery phase of the ongoing trial, where lawyers for the NYT have said the messages are necessary to discover whether ChatGPT did indeed copy verbatim text from them.

OpenAI must now first anonymize the logs, and then submit them to the court, and NYT’s attorneys, seven days later.

Read more: Reuters has the scoop.

Altman declares «Code Red» at OpenAI, plans Gemini 3-beating model

OpenAI has put ads on hold until it can release a Gemini 3-beating model.
All other initiatives are on hold as OpenAI prepares its next model. (Picture: generated)
Competition is heating up in the AI chatbot market, as highlighted in the last weeks, with new, capable models from Google, Anthropic, and a new Codex Max from OpenAI.

This has now caused Altman to delay other initiatives, such as ads, to focus on making a better ChatGPT, paywalled The Information writes, citing an internal memo.

They are apparently planning to release a new reasoning model next week that will be «ahead of Gemini 3.» But this needs a little more polishing on the «experience.»

Just last week, a developer revealed ad code in the latest ChatGPT beta — meaning that their work on ads was fairly advanced and almost ready to ship.

This work is now on the back burner, at least until next week, when OpenAI hopes to reclaim their crown.

Read more: The Information (paywalled), and Reuters. Discussion on r/Singularity.

Three years since the debut of ChatGPT 3 — the first AI chatbot

Little did OpenAI know of how the AI age was about to ignite with their research bot.
ChatGPT was launched on an unsuspecting world some three years ago today, and promtly exploded in users. (Picture: Generated)
If you’re in the USA, today marks the third birthday of the first ChatGPT bot from OpenAI, on November 30, 2022.

It was launched «expecting a small research audience,» OpenAI’s Nick Turley says on x.com, but «within hours, the usage was far beyond anything we’d prepared for.»

Launch page is still live
On the preserved introduction page at OpenAI, it says «we are presenting ChatGPT, and look forward to feedback from users to learn more about the strengths and weaknesses of the model.»

Continue reading “Three years since the debut of ChatGPT 3 — the first AI chatbot”

Developer finds ad references in latest ChatGPT beta

ChatGPT with ads? They could actually pay for more features while offering a revenue stream for OpenAI.
These little lines of code could be a harbinger of ads on the ChatGPT app. (Picture: Tibor Blaho)
This seems unconfirmed by anyone else, but strongly suggests that OpenAI is working on an ads product for their chatbot.

Tibor Blaho discovered the changes in the beta app for Android while doing routine versioning work for his job at AIPRM.com, which integrates tightly with ChatGPT.

He also likes to «share what is coming next with people who care about ChatGPT and Claude news.»

The new code indicates an ads «bazaar,» and a «SearchAd» with a «SearchAdsCarousel.»

Presumably, ads would begin rolling out on the free version of ChatGPT, where it could pay for offering more features. Or there could be affiliate marketing in the shopping links they just announced.

Google is already rolling out ads on their AI products, and Copilot and Perplexity have started using ads on shopping and follow-up-questions, notes Blaho.

Read more: Tibor Blaho’s x.com post, Bleeping Computer, discussion on r/Singularity.

OpenAI can’t use the word «Cameo» inside and outside of Sora

OpenAI is now barred from using the "Cameo" word, which is a major feature in the Sora 2 generator
Sora 2 is barred from making use to the term until December 22, when a new hearing will be held. (Picture: generated)
Sora’s launch wasn’t just about the ability to make realistic short-form videos, but heavily featured the «Cameo» ability.

This lets you create custom characters of friends or yourself and re-use your «Cameo» in different settings.

Not so fast, said the makers of the real «Cameo,» which sells custom-made celebrity videos or greetings. This is their whole business model, and they promptly sued to get their name back.

Now, U.S. District Judge Eumi K. Lee has granted a temporary restraining order on the use of the word by OpenAI — inside the app and elsewhere — until a hearing can be held on whether or not the ban should be made permanent on December 22.

Read more: scoop by CNBC, writeups by Engadget and Gizmodo

Jony Ive and Altman detail new device as something «to take a bite of»

OpenAIs hardware device is in prototype and is less than two years out, Altman says
Delectable, delickable, Ive is making the ultimate consumer device for OpenAI. (Picture: OpenAI, screenshot).
Ive and Altman teamed up to build a ChatGPT consumer device some six months ago and seem to have settled on a prototype.

They are looking to make a device that’s «simple and playful,» The Verge reports, and it is rumored to be screen-free and about the size of a smartphone, they say.

Continue reading “Jony Ive and Altman detail new device as something «to take a bite of»”

OpenAI introduces GPT-5.1-Codex-Max, defeating Gemini 3 on some benches

Codex-Max reaches parity with Gemini 3, just a day after launch. (Picture: Screenshot, OpenAI)
OpenAI’s new coding model outperforms «state of the art» Gemini 3 from just yesterday, in some select benchmarks — and seems to be on par at SWE-Bench Verified.

— GPT‑5.1-Codex-Max is faster, more intelligent, and more token-efficient at every stage of the development cycle–and a new step towards becoming a reliable coding partner, says OpenAI in their launch post.

It has been observed by the AI lab to work independently on tasks for more than 24 hours, iterating on its implementations and delivering a «successful result.»

Codex-Max is also the first OpenAI model trained in a Windows environment, and will achieve better performance than the previous GPT-5.1-Codex using 30% fewer tokens — meaning it’s cheaper and more efficient.

Read more: The launch post, VentureBeat. Discussion on r/Singularity.

OpenAI launches GPT-5.1 in Instant and Thinking modes

OpenAI is launching "warmer and more emphatic" GPT-5.1.
OpenAI have stopped posting benchmarks for their new models, but they should do better in AIME 2025 and Codeforces. (Picture: OpenAI)
Available for paid tiers today, GPT-5.1 Instant will become the new default model in ChatGPT.

OpenAI are uncharacteristically not posting any benchmarks on their launch page, simply describing both models as «more capable and useful,» and «more intelligent.»

Both models are also supposed to be «warmer and more emphatic.»

People want friendly bots
— We heard clearly from users that great AI should not only be smart, but also enjoyable to talk to, OpenAI writes.

Continue reading “OpenAI launches GPT-5.1 in Instant and Thinking modes”