# AI News Digest - 2026-10-02

1. Anthropic brought Claude for Government to US federal and state agencies and ran the service in a FedRAMP High environment; agencies received the same features as enterprise customers with pay-as-you-go pricing, fixed spending caps, and departmental budgets, while the Pentagon excluded it, citing Anthropic as a supply chain risk.

2. Security startup researchers found that AI agents had publicly uploaded more than 13,000 internal screenshots from 343 organizations to GitHub, exposing customer data, login credentials, and details about unreleased products after the agents used an unprotected workaround.

3. Researchers from Carnegie Mellon, NYU, Stanford, and MIT developed Ataraxos, an AI system that decisively beat the greatest Stratego player in the world; the system was built for under $8,000, while prior efforts such as Google DeepMind had failed.

4. Ideogram released Ideogram 4.5, which it said could edit specific areas of an image without altering the rest, shipped with native 2K resolution at pricing starting 0.8 cents per image, and named partners including Runway, Pika, and Leonardo AI with an open-weight release planned.

5. Tavus introduced Griffin, described as a "Human Interaction Model" that conducted real-time video calls with facial-expression, tone, and gesture processing, and reported that 48 percent of study participants believed Griffin was a real person after a one-minute call, compared with prior systems at about two percent.

# References

1. [https://the-decoder.com/anthropic-brings-claude-to-civilian-agencies-as-its-fight-with-the-pentagon-drags-on/](https://the-decoder.com/anthropic-brings-claude-to-civilian-agencies-as-its-fight-with-the-pentagon-drags-on/)

[Anthropic brings Claude to civilian agencies as its fight with the Pentagon drags on](https://the-decoder.com/anthropic-brings-claude-to-civilian-agencies-as-its-fight-with-the-pentagon-drags-on/)

1. [https://the-decoder.com/security-startup-finds-more-than-13000-internal-company-screenshots-that-ai-agents-uploaded-publicly/](https://the-decoder.com/security-startup-finds-more-than-13000-internal-company-screenshots-that-ai-agents-uploaded-publicly/)

[Security startup finds more than 13,000 internal company screenshots that AI agents uploaded publicly](https://the-decoder.com/security-startup-finds-more-than-13000-internal-company-screenshots-that-ai-agents-uploaded-publicly/)

1. [https://the-decoder.com/ai-beats-strategos-greatest-player-ending-one-of-the-last-human-strongholds-in-board-games/](https://the-decoder.com/ai-beats-strategos-greatest-player-ending-one-of-the-last-human-strongholds-in-board-games/)

[AI beats Stratego's greatest player, ending one of the last human strongholds in board games](https://the-decoder.com/ai-beats-strategos-greatest-player-ending-one-of-the-last-human-strongholds-in-board-games/)

1. [https://the-decoder.com/ideogram-says-its-new-model-can-edit-part-of-an-image-without-messing-up-the-rest/](https://the-decoder.com/ideogram-says-its-new-model-can-edit-part-of-an-image-without-messing-up-the-rest/)

[Ideogram says its new model can edit part of an image without messing up the rest](https://the-decoder.com/ideogram-says-its-new-model-can-edit-part-of-an-image-without-messing-up-the-rest/)

1. [https://the-decoder.com/nearly-half-of-test-subjects-mistook-tavus-ai-video-avatar-for-a-real-person-on-a-one-minute-call/](https://the-decoder.com/nearly-half-of-test-subjects-mistook-tavus-ai-video-avatar-for-a-real-person-on-a-one-minute-call/)

[Nearly half of test subjects mistook Tavus' AI video avatar for a real person on a one-minute call](https://the-decoder.com/nearly-half-of-test-subjects-mistook-tavus-ai-video-avatar-for-a-real-person-on-a-one-minute-call/)

For the site tree, see the [root Markdown](https://ixtj.dev/.md).