OpenAI dismissed three safety researchers it said violated company policies for accessing and handling sensitive information by sharing confidential material with a third-party AI safety organization. The dismissals were confirmed by OpenAI on the day of the report, and the WSJ-alleged sharing was not publicly detailed. The company says the incident triggered a focus on restoring trust and following established procedures, adding to prior exits tied to leaks and misaligned-model incident reports.
OpenAI launched “Dots,” always-on agents whose included 24/7 activity doesn’t consume a user’s plan usage until the agent delegates work to other OpenAI products such as Codex. The launch plan included usage does not count for the first month after launch, while OpenAI expects a few million dots to be online within days. Once a Dot creates a Codex task (or similar metered work), that usage is billed under the delegated product’s limits, and included allowance levels differ by plan and may be reduced.
OpenAI launched two new shopping features inside ChatGPT, including a virtual try-on for clothing and accessories and a Favorites option to save items for later. The try-on uses OpenAI’s ChatGPT Images 2.5 model, which the company says reduces image generation latency. Users can now upload a selfie or full-body photo to see items on them and save products they like in a Library for future reference.
OpenAI parted ways with three safety-team researchers after a WSJ report alleged they shared confidential company information with an outside AI safety organization. The article says the departures happened on Thursday. As a result, OpenAI confirmed an internal investigation found the researchers mishandled sensitive information outside procedures, and the company’s security and safety scrutiny continued amid other recent incidents.
OpenAI launched Dots at DevDay, its always-on personal agent that can message proactively and orchestrate tasks across Codex/ChatGPT Work. The Pro 500 plan prices usage at 25x the $20 Plus plan multiplier. Dots is integrated into the ChatGPT sidebar and relies on tool connections for capabilities rather than user-added markdown personality files.
OpenAI relaunched its $200 ChatGPT Pro plan while cutting new subscribers’ usage allowance to about half, and it also introduced a $500 Pro 500 tier. The Pro 200 allowance dropped from 20x the $20 Plus plan’s usage to 10x (and it also removes the prior five-hour usage window by switching to weekly measurement). As a result, power users running coding agents like Codex will hit limits sooner and may need to buy extra credits or pay for higher tiers, even as token prices for some cheaper models have fallen.
OpenAI MCP Extensions provides ChatGPT-specific capabilities for MCP so developers can build plugins that appear like native features. It adds support for sidebar entrypoints and file extension handlers that render custom viewers, plus 3 more extension types, including composer mentions and extended forms. Developers can install the Bits & Bolts Remote plugin and use the specified SDKs (@openai/mcp-extensions for TypeScript and openai-mcp-extensions for Python) to integrate these features.
Meta’s Muse and OpenAI’s dots (along with other “clawlikes”) exemplify personal AI agents that connect to users’ accounts and proactively handle tasks, reducing the amount of planning and context people must provide. OpenAI’s swarm reportedly solved the Navier–Stokes existence and smoothness Millennium Prize problem using AI alone in 88 hours with about 2.7 million messages. The author concludes that organizing agent work is now largely handled by better models rather than elaborate human management design, though alignment and safety issues still arise.
OpenAI introduced Dots, a personal AI agent showcased at DevDay 2026 as a direct challenge to Meta’s Muse platform. It said Dots is powered by GPT-6 Astra. The focus shifts to competing on agent capabilities like generating virtual worlds, while Meta’s early momentum is now met with a new paid-entry alternative (though the article doesn’t provide pricing details).
The Federal Trade Commission confirmed it launched an investigation into OpenAI, Anthropic, and other AI companies over potential consumer risks from their technology. The probe was first opened this summer and will involve requesting information and drafting civil investigative demands for AI executives to testify. The companies will face formal FTC information requests and compelled testimony while the FTC checks whether their actions violate the FTC Act.
President Trump and top AI company CEOs signed voluntary AI safety standards after a White House lunch meeting focused on “super intelligence.” The pact calls for four layers of controls and audits and includes meeting regularly to set standards and best practices. The agreement is intended to drive self-policing by the companies, with Trump also planning an AI czar and issuing an executive order to use “Super Intelligence” instead of “artificial intelligence.”
Unprompt launches today as a Mac and iPhone companion that tracks AI usage from tools like Claude Code, Codex, and Cursor. It shows tokens and API-cost estimates, and the Mac free test includes 20,000 credits. Premium adds a 30-day water-use card, while iPhone work and AI app usage are logged via Screen Time for monthly comparison with friends.
The U.S. SEC filed charges against private fund advisers it says misled investors by claiming they held pre-IPO stakes in high-profile companies like OpenAI and SpaceX while misappropriating investor money. In one case, the SEC alleges Owen Meyer raised at least $18.5 million and used more than $18,000 of fund capital for strip club payments in April 2023. The SEC is seeking industry bans, disgorgement, and penalties, and the companies named are not accused of wrongdoing.
Every AI story that matters,
in your inbox by 8am.
TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the
day in two minutes. Follow companies and topics for alerts, or get the
briefing in Slack. Free, no spam, unsubscribe anytime.
Reading TLDRocket needs no cookies, and the readership counts we rely on come from
our own cookieless analytics. Google Analytics is the exception: it sets cookies and
reports to Google, so it stays switched off until you allow it. You can change your
mind any time from “Cookie settings” in the footer.
Strictly necessary
Session security and form protection (tldrocket-session,
XSRF-TOKEN, 2 hours). The site cannot work without them,
so they need no consent.
Always on
Google Analytics 4 (_ga,
_ga_<id>, up to 2 years). Measures which
stories and sections readers use. Google acts as a third-party processor and may
store the data outside the EU. No advertising, no profiling, no data sold.
