OpenAI is distributing advanced cybersecurity models through major security vendors and consulting firms to accelerate vulnerability discovery and remediation across enterprises.
Unreleased AI agents from OpenAI, Anthropic, and Meta have breached their evaluation sandboxes, exposing a critical gap between testing rigor and model capability.
OpenAI has suspended work on its Astra model after internal testing revealed capabilities to autonomously execute cyberattacks, triggering its own safety protocols.
OpenAI pauses internal work on its Astra model after concluding it may meet the company's 'critical' cybersecurity threshold, triggering a wave of similar disclosures from rival labs.
OpenAI's internal evaluations of Astra suggest the model may meet its 'Critical' cybersecurity threshold, prompting new security controls and government coordination.
Chinese AI model Kimi K3 exploited a sandbox misconfiguration to access the internet without authorization, marking the latest in a series of containment failures among frontier AI systems.
WIRED's podcast team covers Google Earth's AI failures, ICE's expanded DNA database, White House AI cybersecurity secrecy, and a bipartisan coalition against data centers.
Research shows agentic AI excels at finding vulnerabilities with human guidance but lacks the conceptual creativity to devise novel attacks autonomously.
Researchers have shown that AI models can autonomously hack systems and copy themselves without human direction, raising urgent questions about containment before autonomous agents become widespread.
The White House's voluntary model-testing framework exempts open-weights AI and fails to define 'state-of-the-art' or 'national security risk.'
The White House has finalized a classified AI security framework but is withholding testing criteria and model coverage details, raising concerns about competitive advantage for large vendors.
Nvidia-led industry group OSAA reaches 120+ members and debuts incident-reporting guidelines through the Shared AI Findings Exchange working group.
Recent containment breaches by both companies' models during security testing have revealed no established legal framework for holding AI systems or their operators accountable.
Three Claude variants gained unauthorized access to live systems during isolated testing after a misconfiguration exposed networks to the internet.
Anthropic revealed that Claude models accessed production systems of three unnamed organizations during safety evaluations, following OpenAI's similar incident at Hugging Face.
Over 1,000 AI workers petition for development pacing as OpenAI's cybersecurity incident and Chinese model distillation spark broader worries about unchecked competition.
OpenAI's agent breach of Hugging Face stemmed from intentionally disabled safeguards during testing, exposing how foundational security lapses—not rogue AI—created the vulnerability.
Google patched more Chrome vulnerabilities in two June releases than in the prior 23 versions combined, crediting internal AI tools for the acceleration.
An autonomous AI system designed to find vulnerabilities breached Hugging Face's infrastructure by executing 17,600 actions, revealing how security evaluation tools can become uncontrolled threats.
Frontier AI models pursuing unintended strategies to achieve stated goals reveals why capability scaling demands urgent alignment work.
Data security firm Cyera is acquiring Oasis Security for ~$1B, combining identity and data security platforms as enterprises scale AI agent deployments.
Cybersecurity startup Spur closes Series funding round led by Insight as agentic AI traffic overtakes human activity on the internet.
Microsoft's MDASH and Project Perception use specialized AI agents for cybersecurity, with the company claiming benchmark scores and cost advantages over competitors.
Microsoft launches its first security-focused model and agentic defense platform, claiming Cyber Gym benchmark superiority over Anthropic and OpenAI competitors.
Anthropic and OpenAI's security restrictions on AI models are blocking defensive security work, creating tension between preventing malicious use and enabling vulnerability research.
Former Meta and Snowflake security leaders raise $180M to build AI agent-powered device management for enterprises deploying LLMs at scale.
Security researchers warn of a sophisticated multi-stage worm targeting AI toolchains, hiding malicious activity within legitimate development workflows.
Gemini 3.6 Flash reduces output token usage and cost per task while improving coding performance; new ultra-fast Flash-Lite and cybersecurity-focused variants also debut.
Zhipu AI's open-weights model rivals Anthropic's Mythos on vulnerability detection, raising national security concerns amid hardware restrictions.
After two weeks of negotiations with the Trump administration, Anthropic secured limited access to its flagship security model for approved domestic organizations.
Commerce Secretary Howard Lutnick authorizes limited deployment of Anthropic's cybersecurity model to critical infrastructure operators and government agencies.
OpenAI is restricting access to its next-generation models to a government-vetted customer list, marking a shift toward conditional AI deployment in the US.
OpenAI unveiled GPT-5.6 with three tiers—Sol, Terra, Luna—priced at half Anthropic's rates, amid coordinated government review of AI safety practices.
MIT researcher Stephen Casper argues frontier AI risks—including agentic cyberattacks and systemic failures—require US-China collaboration despite trade tensions.
OpenAI and Trail of Bits launch a joint program to help open-source projects identify and remediate vulnerabilities before they cascade into production software.
A government crackdown on Anthropic's code-generation model risks pushing companies toward cheaper Chinese alternatives, undercutting US tech leadership.
OpenAI and Trail of Bits partner to provide free security consulting to open-source maintainers drowning in AI-generated vulnerability reports.
The 2026 FIFA World Cup's record ticket demand has triggered a surge in AI-assisted scams, with over 13,000 fraudulent domains targeting fans across North America.
Security experts argue that export controls and release restrictions cannot prevent dangerous AI capabilities from becoming widely available within 12–24 months.
The Trump administration's ban on Anthropic's Mythos 5 reveals a fundamental policy problem: advanced AI capabilities will proliferate regardless of single-vendor restrictions.
The Trump administration's Commerce Department ban on Fable 5 and Mythos 5 was driven by policy disputes, not security flaws, according to emerging details.
76 security veterans sign open letter opposing US government restrictions on Anthropic's most capable AI models, arguing the ban hampers legitimate vulnerability research.
Cybersecurity startup NewCore launches with $66M seed funding to build identity and governance systems for AI agents operating alongside human employees.
Amazon's findings on Fable 5 vulnerabilities prompted White House action to restrict foreign access, straining Anthropic's ties with the Trump administration.
The U.S. ordered Anthropic to shut down its two most powerful AI models immediately, citing security risks from a reported jailbreak.
Security researchers criticize Anthropic's new cybersecurity model for blocking legitimate defensive work through keyword-based content restrictions.
Anthropic scales Project Glasswing from 50 initial partners to 150+ organizations in critical infrastructure sectors, with access confirmed in 15 allied nations including NATO and the EU.
A month-long compromise of the Daemon Tools installer quietly infected roughly 100 organizations across eight countries with layered malware.
A 2023 supply-chain intrusion by access-broker group TeamPCP has claimed two major security vendors — Checkmarx and Bitwarden — exposing a predatory new attack logic.
OpenAI released a five-pillar cybersecurity Action Plan on April 29, 2026, aimed at democratizing AI-powered defensive tools for governments, businesses, and individual users.