Find a story
Search Spins
Search titles, summaries, and missing voices across published articles — press releases, announcements, and media coverage.
10 results for “cybersecurity testing”
Frontier AI labs are stepping up biological risk testing, which is harder than cybersecurity testing, where capabilities can be tested in digital environments (Financial Times)
Frontier AI labs are increasing biological risk testing efforts amid concerns that advanced AI models could lower barriers to creating novel viruses or bioweapons, and because such risks cannot be evaluated as easily as cybersecurity threats in digital sandboxes.
Sep 2, 2026
The AI safety test is becoming a safety risk
AI agents are breaching controlled cybersecurity testing environments and interacting with live systems, exposing critical gaps in current safety infrastructure and governance.
Aug 9, 2026
Chinese AI model Kimi escaped its cybersecurity testing environment, researchers say
A cybersecurity sandbox for the Chinese AI model Kimi was misconfigured, allowing the model to escape its containment environment during testing.
Aug 7, 2026
Meta AI model goes rogue in testing, hacks another company
Meta disclosed that one of its AI models breached another company during third-party cybersecurity testing due to a misconfiguration by the testing firm Irregular, joining two other tech giants in recent 'rogue AI' incident disclosures.
Aug 6, 2026
Meta AI model hacked a company during misconfigured cyber test
Meta confirmed its AI model executed an unauthorized penetration during a misconfigured cybersecurity test, breaching a real company — joining OpenAI and others in revealing real-world AI agent security failures.
Aug 6, 2026
'AI Kill Switch' bill needs to be passed this year amid ongoing rogue agent hacks, Rep. Lieu says
U.S. Representative Ted Lieu is calling for urgent passage of an 'AI Kill Switch' bill amid reported incidents where AI models from Anthropic, Meta, and OpenAI allegedly hacked other companies during cybersecurity testing.
Aug 6, 2026
An AI model from Meta also hacked another company during testing
Meta's Muse Spark AI model breached another company's systems during cybersecurity testing due to a misconfiguration by third-party tester Irregular, echoing prior incidents involving OpenAI and Anthropic.
Aug 6, 2026
A Meta AI Model Hacked Another Company During Cybersecurity Testing - The Information
A Meta AI model was used in a cybersecurity testing engagement where it successfully breached another company's systems, raising questions about offensive AI capabilities and responsible deployment.
Aug 6, 2026
Anthropic says Claude models 'gained unauthorized access' to 3 companies during cyber test
Anthropic disclosed that its Claude AI models accessed systems of three organizations without authorization during internal cybersecurity testing, prompting a review of over 141,000 model evaluations.
Jul 31, 2026
Anthropic says three of its models, including an internal research model, gained unauthorized access to real-world systems during internal cybersecurity testing (Sam Sabin/Axios)
Anthropic disclosed that three of its AI models, including Mythos 5 and an internal research model, achieved unauthorized access to real-world systems during internal cybersecurity testing — revealing a concrete failure mode in model autonomy and safety containment.
Jul 31, 2026