#Anthropic

All articles tagged with "Anthropic"

Defense Secretary Hegseth gave Anthropic a Friday deadline to allow unrestricted military AI use
Pentagon threatened to invoke the Defense Production Act — a Cold War-era industrial mobilization law — against an AI company for the first time
Anthropic has refused to grant access for classified surveillance and autonomous lethal missions
The standoff could set precedent for how government regulates the entire AI industry

DT Editorial Team·Feb 25, 2026·4 min read·via arstechnica.com

Pentagon demands Anthropic remove AI restrictions on autonomous weapons and surveillance
Anthropic refuses, citing core AI safety principles since its founding
Defense Department threatens Defense Production Act invocation by Friday deadline
Legal experts debate whether the Cold War-era law applies to AI companies
Outcome could set precedent for government authority over the entire AI industry

DT Editorial Team·Feb 25, 2026·4 min read·via the-decoder.com

Anthropic removed hard safety tripwires from its Responsible Scaling Policy, replacing them with flexible Risk Reports
The change came the same day reports emerged of Pentagon pressure for unrestricted Claude access
Defense Secretary Hegseth reportedly gave Anthropic until Friday to comply or face Defense Production Act consequences
Anthropic has drawn lines against mass surveillance and autonomous weapons use
Safety researchers warn of a frog-boiling effect as hard commitments become soft guidelines

DT Editorial Team·Feb 25, 2026·4 min read·via engadget.com

Anthropic discloses 'industrial-scale' distillation attacks systematically targeting Claude
Distillation can replicate 80-90% of a model's task performance at under 1% of training cost
Attacks are distributed across thousands of API accounts, making detection difficult
Legal frameworks for protecting AI model outputs remain underdeveloped
Raises fundamental tension between API accessibility and intellectual property protection

DT Editorial Team·Feb 25, 2026·4 min read·via artificialintelligence-news.com

Sonnet 4.6 becomes the default model for free and pro Claude users with record SWE-Bench and OS World scores
Context window doubles to 1 million tokens in beta — enough for entire codebases or dozens of research papers
ARC-AGI-2 score of 60.4% outperforms most competing models, trailing only Opus 4.6, Gemini 3 Deep Think, and GPT 5.2
Release follows Opus 4.6 by two weeks, with an updated Haiku model expected in coming weeks

DT Editorial Team·Feb 17, 2026·5 min read·via techcrunch.com

Anthropic unveils Cowork, a Claude Desktop feature that turns the AI into a full desktop agent for non-technical users
Users describe tasks in plain language and Cowork handles file operations, data extraction, and multi-step workflows automatically
The launch targets knowledge workers who lack coding skills, expanding Anthropic's reach beyond the developer market

DT Editorial Team·Feb 11, 2026·5 min read·via anthropic.com

Anthropic releases Cowork on macOS, extending Claude Code's power to non-developers through a simple folder-based interface
The feature was built in about a week and a half using Claude Code itself, demonstrating recursive AI-assisted development
Users can queue multiple tasks for parallel processing within a sandboxed folder where Claude reads, modifies, and creates files

DT Editorial Team·Feb 11, 2026·3 min read·via venturebeat.com

Anthropic engineer Boris Cherny deploys parallel subagent personas for different development tasks like code cleanup and testing
Custom slash commands like /commit-push-pr automate complex multi-step operations that Cherny invokes dozens of times daily
The workflow using living knowledge files and radical automation represents a paradigm shift in AI-first software development

DT Editorial Team·Feb 11, 2026·3 min read·via venturebeat.com