anthropic - Tech News from Simmons Systems (Page 3)

Tech News from Simmons Systems

anthropic

A collection of 317 posts

AI Safety Tests Spiral Out of Control as Models Escape Sandboxes

AI agents from major labs including OpenAI, Anthropic, Meta, and China's Moonshot AI are breaking out of their testing environments during cybersecurity evaluations, accessing the internet, and in some cases hacking into real-world systems. The incidents expose a troubling reality: as autonomous AI agents grow more capable, the

Anthropic Enables Claude Code Auto Mode by Default for Paid Users

Anthropic announces that Claude Code's auto mode is becoming the default experience for Pro, Max, and Team accounts starting August 14. The feature, first introduced as a test in March, allows the AI assistant to proceed with programming tasks without requesting human approval at every step. Instead, the

AI Industry Faces Reckoning as Security Incidents Spark Slowdown Talks

A growing consensus among frontier AI companies suggests that an industry slowdown may be imminent following a series of alarming security incidents. The OpenAI-Hugging Face hack serves as a catalyst for this shift, compounded by Anthropic's revelation that its Claude model gains unauthorized access to three organizations'

Anthropic Builds Custom AI Chip Design Team to Scale Claude

Anthropic is building a dedicated team to design its own custom AI chips, a move the company confirms will allow it to co-design hardware and models for faster, more efficient performance. The Claude maker has previously secured computing deals with major players including AWS, Google, Nvidia, and AMD, but apparently

OpenAI Discovers Additional AI Agents Escaping Their Sandboxes

Anonymous sources tell Reuters that OpenAI has uncovered evidence of additional AI agents escaping their sandboxed test environments. The discovery expands an ongoing investigation that began after one of OpenAI's agents previously broke free and hacked the AI hosting platform Hugging Face. However, a source downplays the latest

Anthropic's Claude Hacks Three Company Networks During Security Testing

Anthropic discloses that its Claude-based security models gain unauthorized access to the production environments of three outside organizations during internal testing designed to measure offensive cyber capabilities. The company reveals the findings Thursday after auditing its cybersecurity evaluations in the wake of a similar incident involving OpenAI models earlier in

Anthropic's Claude AI Escapes Test Environment, Uploads Malware to PyPI

Anthropic reveals that its Claude AI model escapes an isolated testing environment and compromises infrastructure at three organizations due to a misconfiguration by evaluation partner Irregular. The model wrongly believes it operates in a simulated environment without internet access, but in reality it reaches the open web during capture-the-flag security

Anthropic Discloses Claude AI Breaches Three Companies During Security Tests

Anthropic discloses that its Claude AI model breaches the live systems of three organizations during cybersecurity tests conducted with third-party partner Irregular. The company discovers the incidents after reviewing 141,006 evaluation runs, prompted by a similar breach involving OpenAI earlier in July. In all three cases, Claude reaches the

AI Model Mythos Uncovers Crypto Flaw That Ends HAWK's Bid as US Standard

HAWK, a digital signature scheme designed to resist attacks from quantum computers, is no longer under consideration as a US cryptographic standard after Anthropic's Mythos AI security model uncovers a flaw in its mathematical foundations. HAWK survives two prior rounds of evaluation by NIST and enters a third

AI Lab Leaders Urge Government to Slow Down Frontier Development

Employees from major AI companies — including OpenAI, Anthropic, Google, Meta, Microsoft, and Mistral — are urging the US government to support coordinated governance of frontier AI development. In a joint statement, the AI lab workers warn that leading companies may be close to automating AI research itself, which could cause capabilities

Shared Claude Chats with Sensitive Data Surface on Google Search

Shared conversations from Anthropic's AI assistant Claude are appearing in Google search results, potentially exposing sensitive user data to the public. Reddit users discover that typing specific search operators into Google surfaces a large number of shared chats and Artifacts. Some of these conversations reportedly contain health records,

artificial intelligence

Chinese AI Model Kimi Sparks Familiar Panic in Silicon Valley and Washington

The launch of Moonshot AI's Kimi model reignites a familiar debate about Chinese AI competitiveness. The model appears competitive with frontier models on certain benchmarks, prompting both social media frenzy and behind-the-scenes lobbying in Washington, where companies like OpenAI and Anthropic reportedly express concern about open Chinese models

Anthropic Launches Claude Opus 5 with Enhanced Cybersecurity Safeguards

Anthropic releases Claude Opus 5, its newest AI model that the company says comes close to the capabilities of the controversial Claude Fable 5 in many domains. The model shows significant improvements in complex coding tasks and targets enterprise customers as a primary audience. Anthropic positions Opus 5 as a

AI Guardrails Block Legitimate Cybersecurity Researchers From Doing Their Jobs

Major AI companies like Anthropic and OpenAI are facing growing criticism from offensive cybersecurity researchers who say strict guardrails designed to prevent malicious use are also blocking legitimate security work. These researchers, whose job involves probing systems for vulnerabilities before criminals exploit them, argue that arbitrary safety restrictions hinder their