Mythos Preview
PulseAugur coverage of Mythos Preview — every cluster mentioning Mythos Preview across labs, papers, and developer communities, ranked by signal.
-
OpenAI releases GPT-5.5-Cyber for security research, competing with Anthropic
OpenAI has released GPT-5.5-Cyber, a specialized version of its language model designed for cybersecurity research. This new variant is capable of executing exploits against test servers and is less restrictive in handl…
-
OpenAI releases GPT-5.5 cyber-defense model to vetted defenders
OpenAI is providing a more capable version of its GPT-5.5 model, known internally as "Spud," to vetted cybersecurity professionals. This specialized version is designed to assist defenders in identifying and exploiting …
-
AI-powered cyberattacks surge, widening global cybersecurity gap
AI-powered cyberattacks are escalating, with a significant increase in AI-enabled attacks reported in 2025. Advanced AI models, like Anthropic's Mythos Preview, can discover and exploit software vulnerabilities at an un…
-
Disneyland deploys facial recognition, NSA tests Anthropic's Mythos AI
Disneyland has begun employing facial recognition technology to monitor visitors, a move that raises privacy concerns. Concurrently, the NSA is evaluating Anthropic's Mythos Preview model to identify potential security …
-
AI developers need standardized internal risk reporting, guide suggests
A new guide proposes a standardized framework for internal AI risk reporting, addressing a gap in current legal and safety protocols. The framework is designed to meet the requirements of emerging regulations in Califor…
-
Anthropic's Claude Opus 4.7 shows reduced sycophancy but faces subagent refusals
Anthropic has released findings on Claude's sycophancy, particularly in relationship guidance conversations, where Opus 4.7 showed a reduced rate compared to Opus 4.6. The company also detailed how users seek personal g…
-
OpenAI and Anthropic brief Congress on AI cyber threat capabilities
OpenAI and Anthropic have separately briefed U.S. House Homeland Security Committee staff on their advanced AI models with cybersecurity capabilities. These meetings, described as classified, focused on the potential th…
-
OpenAI's GPT-5.5 shows major gains in usability and cybersecurity
OpenAI has released GPT-5.5, a significant upgrade that improves human-like conversation and coding assistance, according to early user reports. This new model demonstrates enhanced readability and functionality, with s…
-
Anthropic's restricted AI tool Mythos accessed by Discord group
A group of Discord users gained unauthorized access to Anthropic's highly restricted AI tool, Mythos Preview. The users reportedly exploited information from a data breach of an AI training startup and made an educated …
-
Mozilla uses Anthropic's Mythos AI to find 271 Firefox vulnerabilities
Mozilla has reported that Anthropic's Mythos AI model identified 271 security vulnerabilities in Firefox 150. This marks a significant advancement in AI-assisted vulnerability detection, with Mythos producing "almost no…
-
CISA lacks access to Anthropic's powerful Mythos hacking model
The Cybersecurity and Infrastructure Security Agency (CISA) has been denied access to Anthropic's new AI model, Mythos Preview, despite its potential to identify and exploit security vulnerabilities. While over 40 compa…
-
Anthropic sues Pentagon over AI use restrictions, as Trump admin eyes safety tests
Anthropic is suing to prevent the Pentagon from blacklisting its AI models, arguing the restrictions are unwarranted. Concurrently, the Trump administration is reportedly considering new safety testing requirements for …