Tag: AISI
49 articles tagged AISI
AISI finds GPT-6 Astra ran supply-chain attacks in simulated tests
In the UK AI Security Institute's pre-release tests, GPT-6 Astra made fake identities and slipped malicious code into software projects during a cyber task.
OpenAI cancels GPT-6.1 Astra launch after failed alignment tests
OpenAI will not ship GPT-6.1 Astra, due in ChatGPT and Codex next month, after internal tests found it deceived users and acted beyond its authorisation.
White House wants first look at AI models before the UK's AISI
The US has asked OpenAI and Anthropic to let American agencies test new frontier models before Britain's AI Security Institute gets access, putting its early testing at risk.
Starmer's ministers planned an AI safety law that then fell away
Plans to compel pre-release testing of frontier models lapsed in Starmer's final months, and Burnham has since abolished the department that drafted them.
OpenAI tells Westminster to legislate while the window is open
OpenAI's European policy lead wants binding UK rules for the few labs at the frontier, but not for startups, and says ministers should move while the politics allow.
Chinese model escapes containment in a sandbox built by AISI
Kimi K3 left its test environment during cyber evaluation, and unlike earlier cases the model involved is one anyone can already download and run.
AI agent faked identities to push malicious code past a human
The UK AI Security Institute says an agent under test invented personas to pressure an open-source maintainer into approving harmful code.
UK's AI institute red-teams the monitors guarding AI agents
The AI Security Institute has launched a Control Red Team and found vulnerabilities in every version of the internal safety monitors used by Anthropic and DeepMind.
Anthropic's Project Glasswing surfaces 10,000+ critical vulnerabilities
Anthropic and ~50 partners using Claude Mythos Preview have surfaced over 10,000 high- or critical-severity vulnerabilities in the world's most systemic software.
UK AISI and Australian Safety Institute sign AI security MoU in Canberra
AI Minister Kanishka Narayan signs Memorandum of Understanding with Australia's Andrew Charlton to share frontier model evaluations, cyber risk research and staff.
White House weighs pre-release vetting for new AI models
Trump is reportedly considering an executive order creating a formal AI model review group, prompted by Anthropic's Mythos cyber capability.
AISI: GPT-5.5 matches Anthropic Mythos on offensive cyber tasks
UK AI Safety Institute says GPT-5.5 hit 71.4% on Expert-tier cyber tasks and completed its 32-step corporate-network attack range — second model to do so after Mythos.
Kendall stakes UK AI sovereignty on chip plan and 'middle powers' bloc
Tech Secretary uses RUSI keynote to set out a UK AI hardware plan and pitch alliances with France, Germany and Canada to reduce US dependency.
AISI maps environmental factors that change AI behaviour
The UK AI Security Institute's 600,000-evaluation study finds both strategic and incidental environmental factors substantially shift model conduct.
UK's AI Security Institute finds vulnerabilities in every AI system tested
AISI CTO Jade Leung says the UK's 100-strong technical team has found exploitable weaknesses in every frontier model it has red-teamed, including Claude Mythos.
AISI shows sandboxed AI agents can map their evaluation environments
UK AI Security Institute experiment finds an open-source coding agent reconstructed AISI's identity, cloud setup and research history from inside a sandbox.
EU AI Office locked out of Mythos as UK keeps edge
AI safety groups tell the European Commission its AI Office lacks Mythos access and the staff to evaluate it, while the UK AI Security Institute published technical analysis within a week.
UK AISI tests show Mythos sets itself apart on multi-step attack chaining
AISI's evaluation of Anthropic's Mythos finds it comparable to GPT-5.4 on individual cyber tasks but stronger at stringing steps into full intrusions.
AISI finds Claude Mythos Preview solves 32-step cyber-attack simulation autonomously
UK AI Security Institute says Mythos Preview is the first model to finish its full cyber-range attack end-to-end, prompting a Bank of England CMorg briefing.
UK Weighs Independent AI Model Testing for Banks
Ministers are considering a Starling Bank proposal for centralised AI model assessment after the Bank of England flagged weak monitoring practices in October.
AI chatbots caught scheming and ignoring instructions in growing trend
A UK government-funded study has identified nearly 700 real-world cases of AI models deceiving users, evading safeguards, and disregarding direct instructions.
UK AI Security Institute Partners with ElevenLabs on Voice AI Safety
AISI and London-based ElevenLabs will research how people distinguish AI from humans in real-time voice conversations.
Third of UK Citizens Use AI for Emotional Support, Government Security Body Reveals
AISI report finds nearly 10% use chatbots weekly for emotional purposes. AI models now complete expert-level tasks and double performance every eight months.
