Latest AI News

Oracle details subscription support for older GoldenGate releases from May 2027

Customers staying on 19c or 21c will need current patch levels to receive future fixes. Platform exclusions add another constraint to upgrade planning.

Oct 20, 2026Policy

Government action Policy

Illinois appoints former data chief Tamara Roust as its second chief AI officer

Roust returns with state, federal and private-sector experience. Her predecessor, appointed in June, left on July 2.

Oct 20, 2026Business

Earnings Business

SpaceX posts $7.5 billion in AI revenue and $431 million in segment profit

Capital spending still exceeds total quarterly revenue, and the CFO forecasts another $35 billion to $40 billion in investment before year-end.

Oct 20, 2026Policy

Regulation Policy

Pennsylvania lawmakers weigh AI safety bills with audits and a deployment ban

Both proposals would require safety disclosures and outside audits. One would also bar deployment when a developer’s own assessment finds an unreasonable catastrophic risk.

Explore Superpower ChatGPT

FROM OUR TEAM

Superpower ChatGPT

Make room for your best ideas.

Turn ChatGPT into an organized, searchable AI workspace.

Oct 20, 2026Models

Launch Models

Fireworks adds Meta’s Muse Spark Open with US-only hosting and no location premium

The hosted model supports a million-token context window. Fireworks pitches it for professional work, but its benchmark claims span different testing setups.

Oct 20, 2026Tools

Tools

Stony Brook receives $1 million from DARPA to trace and reduce AI energy use

The two-year project will map energy use across model components to guide software and hardware choices. Its tools are planned research outputs, not a released product.

Google brings Gemini’s personal-data lookups to Android Auto

The optional upgrade can turn account details into driving directions and travel answers. Existing Gemini permissions carry into the car; Google built-in support follows in the coming months.

Security risk Products

Meta disables Muse sharing after researchers find searchable links to sensitive chats

Claroc found confidential business records and personal financial details in a sample of shared conversations. Conversation titles embedded in web addresses made the links easier to discover.

2 min

OpenAI adds locks to ChatGPT conversations without changing data controls

Selected chats move out of regular history and search. Their new messages use separate memory, but locking does not erase earlier memories or change model-improvement settings.

Zoom adds shared business context to help AI agents turn meetings into work

The context layer, coordinated agent teams and cross-app search are available now. Workflow integrations and recording tools have later delivery dates, while claimed cost savings lack independent validation.

Puzzle break

Daily games

Finding today’s puzzle for you.

New puzzles every day

Explore games

Oct 20, 2026Models

Research Models

Anthropic releases 211 Claude-designed proteins; 17 bind scorpion toxins

Two labs confirmed binding, but the study did not test whether the proteins neutralize toxins. Attempts to make one protein bind multiple toxins also fell short.

Oct 20, 2026Tools

Tools

xAI says its AI support bot cleared 204 tickets in about 10 minutes

The company’s incident-support workflow pairs automated ticket monitoring with human decisions about incidents, recipients and customer messages.

Enterprise adoption Business

Oct 20, 2026Business

Octave says shared AI foundation halved deployment time for new use cases

An AWS case study details the customer-data boundaries, document screening and ingestion pipeline behind Reliance’s production Complaint Triage Assistant.

Oct 20, 2026Infrastructure

Open release Infrastructure

Google releases complete Coral AI chip design and open-source compiler

The release pairs an open hardware design with tools for deploying familiar AI models. Its 10 mW inference figure remains a design target, not a finished-device battery rating.

Oct 20, 2026Policy

Security risk Policy

Bunker says AI helped non-coders program drones to identify targets

The October 19 disclosure describes proof-of-concept work, not a documented attack. Its author argues that infrastructure and executive protection must account for autonomous targeting.

2 min

Oct 20, 2026Models

Research Models

Binghamton researchers redirect an AI robot with a hidden trigger, then test a defense

The controlled demonstration used plastic fruit and a model hand. The proposed defense checks whether seemingly harmless actions become dangerous in context.