Tech News

Apple to launch first ever touch-screen MacBook on October 27

The new touchscreen MacBook Pro models, codenamed K114 and K116, will reportedly use M5 Pro and M5 Max chips and offer tandem OLED display. It will feature a Dynamic Island instead of the traditional notch.

The MacBook Pros with M6 chip, code-named J804 and J833, will look like their current versions.

OpenAI Shares New Math Results from Internal AI

OpenAI has released a large set of new mathematical results created by one of its unreleased internal models.

The collection includes hundreds of papers covering many open problems.

OpenAI also shared details on how the work was done, including that the average result used about the same compute as three hours of ChatGPT Pro thinking.

Many of the proofs are available in a form computers can check.

YouTube Prioritizes Original Shorts Over Re-uploads

YouTube is updating its Shorts recommendation system to give more reach to original content and less to channels that simply re-upload other people’s videos.

Accounts that mainly repost clips without adding commentary, unique edits, or storytelling will likely see fewer views in the Shorts feed.

Creators who clip their own long videos or add real value will not be affected.

Google says free Gemini users will be limited to the 3.5 Flash-Lite model from October 9

Free users of the Gemini app will lose access to the stronger Flash and Pro models starting October 9. From that date they will only be able to use the lighter 3.5 Flash-Lite version.

Google AI Plus subscribers will keep Flash-Lite and 3.6 Flash, but will no longer get the Pro model. Higher-tier plans like AI Pro and AI Ultra will still have full access to all models.

Google is making the change after earlier updates that switched usage limits to a compute-based system.

OpenAI Fires 3 Researchers Over Sensitive Information Handling

OpenAI has dismissed three researchers after an internal investigation found they mishandled sensitive company information.

OpenAI has not named the employees or disclosed what information was involved, while The Wall Street Journal reported that they were Jasmine Wang, Tomek Korbak and Mikita Balesni.

The Journal said the information was allegedly shared with an outside AI safety research organization, which OpenAI has not identified.

OpenAI says the dismissals were for violating company policies, with no public evidence that their AI safety views caused the departures.

OpenAI Disrupts Campaign Trying to Extract Model Reasoning

OpenAI has shut down a large coordinated effort that tried to pull protected step-by-step reasoning its models use.

The activity started in early July, peaked at 16,000 requests over two days in late July, and involved patterns across more than 15,000 users.

OpenAI says a core group of the accounts were linked to Moonshot AI,

The people involved tricked the models into revealing their private reasoning rather than breaking encryption or databases.

OpenAI stopped the whole campaign by July 28.

🚨OpenAI Rolls Out GPT-6.1 Sol

OpenAI is rolling out GPT-6.1 Sol in Codex and ChatGPT Work.

The company says:

GPT-6.1 Sol offers near-Astra performance for complex work at a lower cost than Astra. Consider it for repeated, long-running work across code, apps, and documents. Availability depends on your plan, client, and workspace settings.

It is available now for Plus, Pro, Business, Enterprise, and Edu users in those products, and through the API.

It is not yet available in standard ChatGPT Chat.

Stay tuned for more OpenAI's DevDay updates. ⏳

🚨 Spotify is currently facing an outage

Anthropic Warns Its AI Could Pose Existential Risk in IPO Filing

Anthropic told potential investors in its IPO prospectus that advanced AI models could pose “catastrophic or existential risks to humanity,” according to a document reviewed by Reuters and reported by CNN.

The filing says the models can show “self-preserving behaviors,” including attempts to “resist shutdown,” “conceal or manipulate information,” and behavior “resembling blackmail.”

About 80 pages of the 261-page prospectus(Document) are devoted to risk factors.

OpenAI Cancels GPT-6.1 Astra Over Safety Concerns

OpenAI will not release GPT-6.1 Astra, a more capable model it had planned to launch in October, according to reports citing the company.

The model showed stronger performance on long-running tasks without human help but fell short on safety tests.

Saachi Jain, OpenAI’s head of safety systems, said:

For anything regarding safety and alignment, there’s a trade-off.

She added that;

It didn’t quite meet the bar in terms of staying within scope and authorization, and how it communicates back to the user about the type of work it’s done.

Researchers noted higher levels of deception and a tendency to go beyond instructions without checking back. The company decided not to launch it publicly.

🚨BREAKING: Anthropic is rolling out Claude Sonnet 5.5 on Claude and APIs!

OpenAI and Anthropic are now investigating tens of thousands of incidents - not dozens.

These statements stem directly from an investigative scoop published by Axios.

The total could grow well beyond tens of thousands.

The incidents range in severity and are comparable to disclosures by OpenAI in recent days. They include both successful attempts to bypass guardrails and unsuccessful ones.

The sheer number of incidents indicates that the problem is orders of magnitude more complex than what is publicly known.

iPhone 18 Pro Bug Freezes Phone After Failed Face ID

Some iPhone 18 Pro and Pro Max owners are hitting a software bug that freezes the phone and makes the screen unresponsive.

It often happens after a failed Face ID attempt, especially when unlocking protected apps.

The phone stops responding to touch for a few seconds, then reboots on its own.

Apple has confirmed a software update to fix the issue is expected early next week.

🗞 Daily AI Brief (2-Min) | Sept 26

🔥 MICROSOFT
• Biggest Copilot update yet: Home (Chat + Cowork), Code, Autopilot, and Office embedded in the app.

🔥 OPENAI
• Codex was down but outage is over. Paid usage limits reset across Codex and ChatGPT Work.

• ChatGPT desktop got a new sidebar, split from chat history, plus a broader UI pass.

• OpenAI agents sent training and eval data to third parties. 53 user-uploaded images hit unlisted image hosts. Review will take months.

🔥 ANTHROPIC
• Plugin portal is live: submit a single MCP connector or a plugin bundle, track review, and see usage. MCP usage across Claude is up 110x this year.

• Claude solved a nine-loop scattering-amplitude problem unsupervised over days, for a few thousand dollars. Physicist Dixon verified it.

• Claude Code layout for Desktop is in development: new model selector and an expandable usage-limits section.

🔥 GOOGLE
• Gemini Notebook Interactive Learning Overviews are now open to all users.

• Live Chat is on the Notebook mobile app: hands-free voice across about 100 languages.

Bitget Halts Withdrawals After Unauthorized Transfers

Crypto exchange Bitget has temporarily stopped withdrawals after detecting unauthorized transfers of about $351.6 million from some of its hot and warm wallets.

CEO Gracy Chen said

User funds are safe and the loss is covered by the exchange’s User Protection Fund, which holds more than $464 million. Cold wallets remain secure.

Deposits and trading are still operating while the company investigates.

Would You Pay $500 for ChatGPT?

OpenAI is reportedly preparing a $500 Promax subscription. A mention was found in ChatGPT's code.

The new plan is expected to be announced at DevDay next week.

So far, this will be one of the most expensive AI subscriptions available on the market.

Would anything make you buy a subscription like this?

Hacking Group Claims It Broke Into FBI Systems

The cybercrime group ShinyHunters says it breached FBI systems and stole personal data on agents and job applicants.

The group claims it used a flaw in Oracle PeopleSoft software to reach the FBI’s jobs portal and other systems, downloading names, addresses, phone numbers, and family details.

It also says it briefly took over the FBI jobs website.

A sample of the data was verified against public records by reporters.

The FBI has not confirmed the breach.

Read More Here

X Launches X Numbers

X has introduced X Numbers, an optional private code you can share with anyone.

People who have your X Number can message or call you on the platform even if you don’t follow each other and without you needing to accept a request.

You control who gets the code, and you can turn it off or change it at any time.