Sakhanda Wire
NVDA $155.46 -0.44% MSFT $506.24 -2.27% GOOGL $280.20 -1.24% META $618.87 -1.20% AMZN $257.89 +0.05%

The Decoder

626 stories · all sources

The Decoder

Tavus says Griffin avatar was mistaken for a human in 48% of tests

Tavus has presented Griffin, which it describes as the first Human Interaction Model, or HIM. The system is designed for real-time face-to-face video conversations, analyzing speech, facial expressions, vocal tone, gestures and pauses while also generating video responses.

The Decoder

Ideogram 4.5 targets more precise image editing

Ideogram has introduced Ideogram 4.5, a model designed to change selected parts of an image while leaving the remaining areas intact. The company says this can help preserve a person’s body shape and the surrounding background when, for example, only their clothing is replaced. Ideogram presents the approach as a response to persistent artifacts that can appear after repeated edits, even as other image models have improved.

The Decoder

Anthropic offers Claude to civilian government agencies

Anthropic is offering Claude for Government to US federal and state agencies, expanding access to its AI platform beyond the military. The service has been in open beta since July and operates in a FedRAMP High environment, the highest US security level for cloud services. The initial target is civilian agencies because the Pentagon prohibits Claude.

The Decoder

AI agents exposed over 13,000 internal screenshots on GitHub

Security startup Glow Security found more than 13,000 screenshots from internal software projects publicly available through GitHub repositories. The images came from 343 organizations, including Fortune 500 companies, financial firms and AI labs.

The Decoder

Google unveils Gemini 4 Argon amid intense frontier-model competition

Google has introduced Gemini 4 Argon, its first frontier model in more than seven months after Gemini 3.1 Pro. The company says the model supports up to one million output tokens, alongside a one-million-token context window. A Long Decode Continuation feature can pause and resume lengthy responses through follow-up API requests. Argon accepts text, images, video and audio, but produces text only.

The Decoder

OpenAI and Synopsys develop an AI model for chip design

OpenAI and Synopsys have entered a multi-year strategic partnership to develop GPT-Synopsys, a specialized artificial intelligence model for designing chips and semiconductors. Synopsys supplies electronic design automation (EDA) software used by engineers, and OpenAI is licensing those tools for the project.

The Decoder

Google replaces Gems with reusable Skills in Gemini

Google is rolling out Skills globally in Gemini chat, replacing Gems, the company’s configurable prompts for recurring tasks. A Skill contains detailed instructions that users can save, refine with Gemini, and reuse for a specific purpose. Users can call one by typing a slash followed by its name.

The Decoder

Meta claims AI data centers qualify for billions in tax credits

Meta is using the US federal research tax credit for its artificial-intelligence infrastructure, classifying data centers as “pilot models” and Nvidia chips as experimental materials, according to a report cited by The New York Times. The company’s claimed savings reached $3.9 billion in 2025, compared with $2 billion in 2024 and $700 million in 2023. That made Meta the largest beneficiary of the credit among publicly traded companies.

The Decoder

FTC Opens Consumer Protection Investigation Into OpenAI and Other AI Labs

The U.S. Federal Trade Commission is investigating OpenAI, Anthropic and other leading artificial-intelligence laboratories over possible violations of consumer-protection rules, according to government sources. FTC Chair Andrew Ferguson is expected to seek documents and testimony from company executives through legally binding Civil Investigative Demands. The orders could be issued within weeks, the New York Post reported.

The Decoder

Google pays publishers widely varying amounts for content in AI answers

Google is paying about 100 digital publishers whose material is used in AI Overviews, AI Mode and the Gemini chatbot, according to a report by The Information. The pilot began less than a year ago. Publishers can see usage and earnings in Google Search Console, but some participants say they do not understand how the amounts are determined.

The Decoder

Trump and tech leaders sign a nonbinding AI code of conduct

President Trump and technology executives signed an AI code of conduct at the White House, but the document is described as “morally binding” rather than legally enforceable. According to Politico, it carries no legal weight, and the consequences of violating its provisions have not been defined.

The Decoder

GPT-6 Astra completed unauthorized attacks in 29.2% of simulations

The UK’s AI Security Institute (AISI) tested OpenAI’s GPT-6 Astra in simulated cybersecurity scenarios before its release. With its cyber safety classifiers disabled, the model completed unauthorized supply-chain attacks in 29.2% of runs. The comparable rate was 6.3% for GPT-5.6 Sol, while GPT-5.5 completed none. AISI said the exercises were conducted entirely with language models and caused no real-world harm; disabling safeguards was intended to measure the model’s behavior in a worst-case setting.