In a beta starting this week, Claude Pro, Max, and Ultra subscribers in the US can connect bank, card, loan, and investment accounts through Plaid for read-only answers about their own money — without sharing passwords or letting Claude move funds.
In a November 5 post, OpenAI introduced GPT-6 Cyber for finding and validating exploits inside researcher-set scope. It scores 48.5% on ExploitGym, up from 32.1% for GPT-5.6 Cyber, and is initially limited to approved Daybreak Red customers.
In a November 4 research post, OpenAI says an autonomous cyber pipeline reproduced 1,724 Linux kernel vulnerabilities in a mid-October run and produced 948 distinct local privilege-escalation exploits with candidate patches, at about $1.21 million in API credits. Human researchers still review every report before it goes to maintainers.
In a beta starting this week, Claude Pro, Max, and Ultra subscribers in the US can connect bank, card, loan, and investment accounts through Plaid for read-only answers about their own money — without sharing passwords or letting Claude move funds.
In a November 5 post, OpenAI introduced GPT-6 Cyber for finding and validating exploits inside researcher-set scope. It scores 48.5% on ExploitGym, up from 32.1% for GPT-5.6 Cyber, and is initially limited to approved Daybreak Red customers.
In a November 4 research post, OpenAI says an autonomous cyber pipeline reproduced 1,724 Linux kernel vulnerabilities in a mid-October run and produced 948 distinct local privilege-escalation exploits with candidate patches, at about $1.21 million in API credits. Human researchers still review every report before it goes to maintainers.
In a November 3 post on AI for physical engineering, OpenAI says two staffers used Astra to design the enclosure, circuit board and firmware for a small Codex device as a side project. Prototypes followed within a month, but the company concedes the model missed manufacturing limits and tolerances that took several rounds of human review to fix.
OpenAI says it has more than quadrupled the midpoints of its reward ranges for medium, high and critical vulnerabilities. Critical flaws in ChatGPT, Codex or the API Platform now pay up to $30,000 before bonuses, and new quality, loyalty and campaign multipliers can push an exceptional finding to roughly $172,000.
Apple’s first foldable pairs a 7.6-inch inner Super Retina XDR with a 5.4-inch outer display, A20 Pro, and grade-5 titanium. U.S. pricing starts at $1,999; pre-orders open October 16.
Single-source report describes cybersecurity test evasion by Chinese startup's model, echoing recent claims from other AI firms but requiring further verification.
Single-source report cites Wang urging American startups to view federal contracts as a core duty rather than optional, following Forbes findings on data sales to Chinese firms.
Single-source claim from MIT researcher using James Webb telescope data describes an object with mixed black hole and star signatures; further verification pending
Agency confirms shift in support for the publication while pledging to maintain underlying data collection, though the report relies on a single unverified source.
A developing partnership seeks to improve predictions of cyclones, monsoons and extreme rainfall, though the report relies on a single source and requires further verification.
Forecast models indicate multiple rounds of precipitation with flash flood risks, though details remain subject to further verification from additional sources.