cybersecurity AI News & Updates

US Government Restores Access to Anthropic's Cyber-Model Mythos 5 for Critical Infrastructure

Following a temporary ban due to easily bypassed guardrails, the Trump administration has partially reversed its stance on Anthropic's powerful cybersecurity model, Mythos 5. Over 100 trusted US government agencies and companies, including their non-American employees, are now permitted to access the model to protect critical infrastructure. Meanwhile, Anthropic continues to work with regulators to resolve restrictions on its other model, Fable 5.

US Government Pressures OpenAI to Restrict GPT 5.6 Launch Over Cyber Safety Fears

The Trump administration has pressured OpenAI to limit the initial rollout of its new GPT 5.6 model to select partners under government oversight due to cyber security concerns. This move mirrors Anthropic's restricted release of Claude Mythos, highlighting growing federal anxiety over frontier models' potential to autonomously exploit software vulnerabilities. OpenAI plans a wider release in a few weeks if the limited deployment goes well.

US Export Ban on Anthropic's Cyber Models Highlights Challenges of AI Control

The U.S. government recently banned Anthropic from exporting its powerful cyber-capable AI models, Fable and Mythos, over national security concerns. This move marks a major test of whether export controls can successfully contain frontier AI systems. However, historical precedents with encryption and spyware suggest that such governmental restrictions are often ineffective and easily bypassed by global actors.

US Export Ban on Anthropic's Fable and Mythos Sparks Outcry from Cybersecurity Experts

A coalition of cybersecurity experts has signed an open letter protesting a U.S. government export ban on Anthropic’s highly advanced Fable and Mythos models. The government issued the restriction due to national security and jailbreak concerns, which prompted Anthropic to suspend global access to these models. Critics argue that blocking these models weakens cyber defense capabilities while global adversaries continue to advance their offensive AI tools.

Managing the Digital Workforce: NewCore Raises $66M to Standardize AI Agent Identities and Security

Cybersecurity startup NewCore has secured $66 million in funding to develop an identity and governance platform specifically designed to manage and control AI agents at scale within enterprises. As companies increasingly deploy AI agents as digital employees, NewCore's platform aims to provide critical safety guardrails, including identity verification, lifecycle controls, and human-in-the-loop authorization tools.

US Export Controls Triggered on Anthropic Models Over Amazon Security and Jailbreak Warnings

Amazon CEO Andy Jassy reportedly flagged security and cyberattack risks associated with Anthropic's Claude Fable 5, prompting the US government to issue export controls. This resulted in Anthropic suspending global access to both its Fable 5 and Mythos 5 models. The situation highlights growing government intervention over model vulnerabilities and potential military or cyber risks.

Cybersecurity Community Criticizes Overly Restrictive Guardrails on Anthropic's Fable

Cybersecurity researchers are criticizing the safety guardrails on Anthropic's newly released Fable model, claiming it overly blocks benign inquiries related to coding and security. When triggered by safety keywords, Fable automatically downgrades the session to an older, less-capable model. While some experts find the limitations frustrating, others acknowledge that conservative boundaries are necessary during the early stages of deploying highly capable cyber-adjacent models.

Anthropic Deploys Claude Mythos to Secure Global Critical Infrastructure

Anthropic is expanding its Project Glasswing initiative, granting 150 organizations across 15 countries access to its advanced Claude Mythos model to identify zero-day software vulnerabilities. The deployment targets high-stakes sectors such as power, water, healthcare, and military alliances like NATO to prevent potentially catastrophic cyberattacks. This expansion occurs alongside similar efforts from competitors like OpenAI's GPT-5.5-Cyber.

Google Cloud Security Concerns Highlight Vulnerabilities in AI Deployments

Google Cloud's COO emphasized the necessity of a platform-centric, agentic defense strategy to counter rapidly accelerating cyber attacks in the AI era. However, investigations revealed a vulnerability where developers incurred massive bills due to unauthorized Gemini API access, compounded by a slow 23-minute key revocation window. This gap underscores the friction between theoretical AI safety paradigms and practical infrastructure implementation.

Trump Administration Postpones AI Security Executive Order Citing Innovation Concerns

President Trump has delayed signing an executive order that would require government evaluation of AI models before public release, citing concerns about hindering U.S. technological leadership over China. The proposed order would have mandated AI companies share advanced models with government agencies 14-90 days before launch, following security concerns raised by recent releases like Anthropic's Mythos and OpenAI's GPT-5.5 Cyber, which can rapidly identify and exploit security vulnerabilities.