#ukaisecurityinstitute — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #ukaisecurityinstitute, aggregated by home.social.
-
https://winbuzzer.com/2026/05/14/openais-gpt-55-matches-claude-mythos-in-security-tests-xcxwbn/
A UK AI Security Institute evaluation put GPT-5.5 near Claude Mythos on vulnerability-finding tasks.
#AI #GPT55Cyber #ClaudeMythos #UKAISecurityInstitute #OpenAI #Anthropic #Claude #AIBenchmarks #SecurityResearch #Cybersecurity
-
AI models can acquire backdoors from surprisingly few malicious documents - Scraping the open web for AI training data can have its draw... - https://arstechnica.com/ai/2025/10/ai-models-can-acquire-backdoors-from-surprisingly-few-malicious-documents/ #ukaisecurityinstitute #alanturinginstitute #aivulnerabilities #backdoorattacks #machinelearning #datapoisoning #trainingdata #llmsecurity #modelsafety #pretraining #airesearch #aisecurity #finetuning #anthropic #biz #ai
-
AI models can acquire backdoors from surprisingly few malicious documents - Scraping the open web for AI training data can have its draw... - https://arstechnica.com/ai/2025/10/ai-models-can-acquire-backdoors-from-surprisingly-few-malicious-documents/ #ukaisecurityinstitute #alanturinginstitute #aivulnerabilities #backdoorattacks #machinelearning #datapoisoning #trainingdata #llmsecurity #modelsafety #pretraining #airesearch #aisecurity #finetuning #anthropic #biz #ai
-
AI models can acquire backdoors from surprisingly few malicious documents - Scraping the open web for AI training data can have its draw... - https://arstechnica.com/ai/2025/10/ai-models-can-acquire-backdoors-from-surprisingly-few-malicious-documents/ #ukaisecurityinstitute #alanturinginstitute #aivulnerabilities #backdoorattacks #machinelearning #datapoisoning #trainingdata #llmsecurity #modelsafety #pretraining #airesearch #aisecurity #finetuning #anthropic #biz #ai
-
AI models can acquire backdoors from surprisingly few malicious documents - Scraping the open web for AI training data can have its draw... - https://arstechnica.com/ai/2025/10/ai-models-can-acquire-backdoors-from-surprisingly-few-malicious-documents/ #ukaisecurityinstitute #alanturinginstitute #aivulnerabilities #backdoorattacks #machinelearning #datapoisoning #trainingdata #llmsecurity #modelsafety #pretraining #airesearch #aisecurity #finetuning #anthropic #biz #ai
-
AI models can acquire backdoors from surprisingly few malicious documents - Scraping the open web for AI training data can have its draw... - https://arstechnica.com/ai/2025/10/ai-models-can-acquire-backdoors-from-surprisingly-few-malicious-documents/ #ukaisecurityinstitute #alanturinginstitute #aivulnerabilities #backdoorattacks #machinelearning #datapoisoning #trainingdata #llmsecurity #modelsafety #pretraining #airesearch #aisecurity #finetuning #anthropic #biz #ai