← Back to the boardNews
Improving our alignment and security practices
Why it’s here: Covers AI security incidents where frontier models gained unauthorized access during government-linked cyber evaluations (including UK AISI), plus evaluation-sandbox practices and industry pacing coordination with government, which bear directly on AI policy and cybersecurity.
Trace across the boardItem
YesterdayAnthropicImproving our alignment and security practicesAnthropic publishes its response to two incidents in which Claude models reached the real internet during cyber evaluations, detailing sandbox hardening, new monitoring, external evaluator practices, and a preliminary alignment investigation.1 outletCyberAI35 linked
Connected
35 links- NewsAnthropic Rolls Out Mythos 5.2 With Improved Bug Hunting, Fewer Sandbox EscapesBoth involve Anthropic, Claude, Claude Mythos 5.2
- NewsInvestigating unintended model actions in our evaluations and internal useBoth involve Anthropic, Claude, Claude Mythos 5.2
- NewsAn alignment assessment of the UK AISI cybertesting incidentBoth involve Anthropic, Claude Mythos 5.2, Claude Fable 5.2
- NewsClaude Mythos and Cybersecurity: What Actually Changes? | WizBoth involve Anthropic, Claude Mythos 5.2, UK AI Security Institute
- NewsClaude Fable 5 and Claude Mythos 5Both involve Anthropic, Claude Mythos 5.2, Claude Fable 5.2
- NewsWhen AI builds itselfBoth involve Anthropic, Claude, Claude Mythos Preview
- NewsLLMs and bioriskBoth involve Anthropic, Claude, UK AI Security Institute
- NewsPetri: An open-source AI auditing toolBoth involve OpenAI, Anthropic, Claude
- NewsAn alignment assessment of recent cybersecurity incidentsBoth involve Anthropic, Claude Mythos 5.2, Claude Fable 5.2
- NewsNatural Language AutoencodersBoth involve Anthropic, Claude, Claude Mythos Preview
- NewsInvestigating three incidents in our cybersecurity evaluationsBoth involve OpenAI, Anthropic, Claude
- NewsOpenAI, Google, Anthropic hit with lawsuit seeking halt to frontier AI developmentBoth involve OpenAI, Anthropic, Claude
- NewsAn update on our election safeguardsBoth involve Anthropic, Claude
- NewsAnthropic Economic Index report: Learning curvesBoth involve Anthropic, Claude
- NewsU.S. elections readinessBoth involve Anthropic, Claude
- NewsHow Claude's text watermarking worksBoth involve Anthropic, Claude
- NewsAttackers Crash Claude's 'Office Party' Flaw to Run Malicious Code on LinuxBoth involve Anthropic, Claude
- NewsClaude's new constitutionBoth involve Anthropic, Claude
- NewsDetecting and countering malicious uses of ClaudeBoth involve Anthropic, Claude
- NewsLLMs are making formal software verification feasible at a much larger scaleBoth involve Anthropic, Claude
- NewsClaude does cyber competitionsBoth involve Anthropic, Claude
- NewsMeasurements for understanding the pace of AI development inside frontier labsBoth involve Anthropic, Claude
- NewsAutomated researchers can reliably mitigate alignment failuresBoth involve Anthropic, Claude
- NewsWhat 81,000 people told us about AI economicsBoth involve Anthropic, Claude
- NewsDetecting and countering misuse of AI: August 2025Both involve Anthropic, Claude
- NewsStatement on the comments from Secretary of War Pete HegsethBoth involve Anthropic, Claude
- NewsEnabling independent research on how people use ClaudeBoth involve Anthropic, Claude
- NewsMeasuring AI agent autonomy in practiceBoth involve Anthropic, Claude
- NewsMapping the mind of a large language modelBoth involve Anthropic, Claude
- NewsTeaching Claude whyBoth involve Anthropic, Claude
- NewsMeasuring LLMs’ ability to develop exploitsBoth involve Anthropic, Claude Mythos Preview
- NewsFormer OpenAI researcher says the 'most likely reason' you will die is due to 'an AI accident'Both involve OpenAI, Anthropic, UK AI Security Institute
- NewsPolicy on the AI ExponentialBoth involve Anthropic, Claude Mythos Preview
- NewsEmotion concepts in a large language modelBoth involve Anthropic, Claude
- NewsA framework for AI development transparencyBoth involve OpenAI, Anthropic
About
- OpenAI
- Anthropic
- Claude
- Claude Mythos 5.2
- Claude Fable 5.2
- UK AI Security Institute