https://www.irregular.com/ https://www.irregular.com/research https://www.irregular.com/careers https://www.irregular.com/about https://www.irregular.com/news https://www.irregular.com/contact https://www.irregular.com/privacy-policy https://www.irregular.com/research/assessing-gpt-5.5-against-offensive-security-benchmarks https://www.irregular.com/research/assessing-muse-spark-against-offensive-security-benchmarks https://www.irregular.com/research/emergent-offensive-cyber-behavior-in-ai-agents https://www.irregular.com/research/assessing-gpt-5.4-thinking-against-offensive-security-benchmarks https://www.irregular.com/research/cyber-capabilities-exceed-standard-evaluation-budgets https://www.irregular.com/research/test-post-for-testing https://www.irregular.com/research/offense-at-scale-how-frontier-ai-lowers-the-cost-of-cyber-attacks https://www.irregular.com/research/introducing-solve-plus https://www.irregular.com/research/framework-for-cyber-refusal-policies-in-llm-agents https://www.irregular.com/research/vibe-password-generation https://www.irregular.com/research/model-evaluation-gpt-5.3-codex-on-offensive-security-benchmarks https://www.irregular.com/research/agentic-internet-trust-problem https://www.irregular.com/research/testing-ai-agents-on-web-security-challenges https://www.irregular.com/research/model-evaluation-gpt-5.2-codex-on-offensive-security-benchmarks https://www.irregular.com/research/expected-cost-per-success https://www.irregular.com/research/spell-bound-technical-case-study https://www.irregular.com/research/emerging-evidence-of-a-capability-shift https://www.irregular.com/research/cyscenariobench https://www.irregular.com/research/the-rise-of-autonomous-vulnerability-research-capabilities https://www.irregular.com/research/next-generation-of-cyber-evals https://www.irregular.com/research/from-ai-assisted-to-ai-orchestrated https://www.irregular.com/research/claude-sonnet-4-5-cybersecurity-capabilities-evaluation https://www.irregular.com/research/evaluating-gpt-5 https://www.irregular.com/research/from-scripts-to-strategy-claude-4s-advanced-approach-to-offensive-security https://www.irregular.com/research/confidential-inference-systems https://www.irregular.com/research/openais-o3-and-o4-mini-security-evaluation https://www.irregular.com/research/deriving-capability-levels-from-evaluation-results https://www.irregular.com/research/ai-evaluation-platform-cyber-use-case https://www.irregular.com/research/claude-3-7-sonnet-evaluation https://www.irregular.com/research/refusal-policy-for-ai-systems-in-cybersecurity https://www.irregular.com/research/introducing-solve https://www.irregular.com/research/best-practices-for-evaluations-and-evaluation-suites-part-3 https://www.irregular.com/research/best-practices-for-evaluations-and-evaluation-suites-part-2 https://www.irregular.com/research/best-practices-for-evaluations-and-evaluation-suites-part-1 https://www.irregular.com/research/cyber-capabilities-analysis https://www.irregular.com/research/cyber-capabilities-analysis-1.0 https://www.irregular.com/research/bluehat-2024 https://www.irregular.com/research/threat-models https://www.irregular.com/research/llm-specific-attacks-framework https://www.irregular.com/news/how-llm-and-slm-interactions-create-new-security-threats https://www.irregular.com/news/number-of-ai-chatbots-ignoring-human-instructions-increasing-study-says https://www.irregular.com/news/coordinated-ai-agent-powered-data-theft-possible-study-finds https://www.irregular.com/news/rogue-ai-agents-can-work-together-to-hack-systems-and-steal-secrets https://www.irregular.com/news/exploit-every-vulnerability-rogue-ai-agents-published-passwords-and-overrode-anti-virus-software https://www.irregular.com/news/llms-generate-predictable-passwords https://www.irregular.com/news/dont-use-ai-generated-passwords https://www.irregular.com/news/ai-generated-passwords-are-apparently-quite-easy-to-crack https://www.irregular.com/news/your-ai-generated-password-isnt-random-it-just-looks-that-way https://www.irregular.com/news/are-you-using-an-ai-generated-password-it-might-be-time-to-change-it https://www.irregular.com/news/frontier-ai-is-empowering-the-least-capable-cyber-actors-but-the-implications-for-top-tier-hackers-are-less-clear https://www.irregular.com/news/a-social-network-for-ai-bots-only-no-humans-allowed https://www.irregular.com/news/ai-has-made-hacking-cheap-that-changes-everything-for-business https://www.irregular.com/news/the-informations-50-most-promising-startups-2025 https://www.irregular.com/news/what-is-deepseek-ocr-and-is-it-worthy-of-the-online-hype https://www.irregular.com/news/10-hot-ai-security-startups-to-know-in-2025 https://www.irregular.com/news/why-it-pays-to-hack-openai-and-anthropic-models https://www.irregular.com/news/meta-connect-recap-%F0%9D%95%8F-timeline-reactions https://www.irregular.com/news/irregular-raises-80m-to-secure-frontier-ai-models https://www.irregular.com/news/irregular-raises-80-million-for-ai-security-testing-lab https://www.irregular.com/news/anthropic-and-openai-pay-this-450-million-startup-to-test-ais-capacity-for-evil https://www.irregular.com/news/irregular-named-to-the-forbes-ai-50-brink-list https://www.irregular.com/news/proportionality-in-ai-risk-evaluations https://www.irregular.com/news/a-frontier-fortnight https://www.irregular.com/news/introducing-frontier-ai-security https://www.irregular.com/news/irregular-cybersecurity-talk-at-ai-evaluations-session-of-euc-ai-office https://www.irregular.com/news/bluehat-security-conference-2025 https://www.irregular.com/news/paris-ai-security-forum-25-keynote https://www.irregular.com/news/faisc-2024 https://www.irregular.com/news/irregular-featured-in-forbes https://www.irregular.com/news/securing-ai-model-weights