Anthropic expands restricted AI security model access to 150 more organizations

- Anthropic expanded Project Glasswing on Tuesday, adding about 150 organizations in more than 15 countries to its restricted AI security testing program.
- The company said its Claude Mythos Preview offensive-security model has surfaced more than 10,000 high- or critical-severity vulnerabilities since April, highlighting rapid gains in AI-assisted defense tools.
- Anthropic warned that comparable Mythos-class models could emerge within six to 12 months, a timeline that underscores the short gap between frontier AI capabilities and broader misuse risk.
- The expansion lands amid a new U.S. executive order seeking voluntary pre-release access for covered frontier models, linking the story to the federal AI safety debate.
Sources & Citations
1 sourceMore Stories
DeepSeek reportedly seeks 6 billion euro funding round amid AI race
• Chinese AI lab DeepSeek is reported to be preparing a 6 billion euro external funding round as it tries to stay competitive in the global model race. • The fundraising would be one of the largest recent AI capital raises, reflecting continued investor appetite for frontier-model developers even as costs and compute needs surge.
Read original · techzine.euPromoting Advanced Artificial Intelligence Innovation and Security – The White House
• President Biden issued an executive order aimed at promoting innovation and enhancing security within the field of advanced artificial intelligence. • The Director of OMB, alongside the National Cyber Director and CISA, must determine within 30 days if existing Federal grant programs can fund advanced AI vulnerability detection.
Read original · whitehouse.gov
White HouseEU's AI Data Center Plans Stumble Due to Delays, Funding Issues
• The European Union's €20 billion ($23.3 billion) investment plan to establish five massive AI data centers is currently floundering due to funding issues and significant delays. • In Germany, major entities including Deutsche Telekom and the Schwarz Group (owner of Lidl) have expressed interest in leading consortium bids to develop one of the planned gigafactories.
Read original · financialpost.comThe Responsible AI Conference 2026 (London): Everything You Need to Know - AI Expert Magazine - premier publication dedicated to AI
• The Responsible AI Conference 2026 will take place as a one-day, practitioner-led event in London, UK, on 23 June 2026. • The event focuses on operational and compliance challenges rather than technical breakthroughs, featuring speakers from high-scrutiny organizations including HSBC, JPMorgan, BBC, and Channel 4.
Read original · aiexpertmagazine.comStrengthening Europe’s tech sovereignty - European Commission
• The European Commission has announced a strategic plan to strengthen Europe's tech sovereignty and establish the region as a global leader in artificial intelligence. • The initiative focuses on increasing digital self-reliance by enhancing capacities in semiconductors, AI, cloud computing, and open-source technologies.
Read original · commission.europa.euEurope is at risk of sleepwalking into AI armageddon
• The European AI ecosystem is currently experiencing a period of extreme hype, characterized by stratospheric valuations and massive funding rounds as of 2026. • Experts warn that "nice-to-have" startups are particularly vulnerable to a potential collapse due to the escalating costs of implementing and maintaining AI technology.
Read original · sifted.eu
SiftedChina robotics firms line up IPOs to pitch next phase of AI
• A surge of Chinese robotics firms are preparing for initial public offerings (IPOs) to fund the next phase of artificial intelligence development. • In Hong Kong alone, at least 46 robotics-related companies are currently in the IPO pipeline, representing over 10% of all applicants.
Read original · straitstimes.comAre Banks Ready to Stand Behind Their AI?
• Leo Li of Ant Digital Technologies asserts that Southeast Asian banks' adoption of AI depends on trust established through tangible outcomes rather than marketing slogans. • Regional regulators are actively addressing AI risks, with Bank Negara Malaysia publishing a discussion paper in 2025 to gather feedback on financial sector AI regulations.
Read original · fintechnews.sg
FintechNewsSGBanks in Southeast Asia Want AI, But Can They Trust It?
• Southeast Asian banks are eager to adopt AI, but Leo Li of Ant Digital Technologies emphasizes that adoption depends on trust built through tangible outcomes rather than marketing slogans. • To establish this trust, regulators are taking proactive steps, such as Bank Negara Malaysia publishing a 2025 discussion paper to gather feedback on AI's regulatory and developmental approach.
Read original · fintechnews.sg
FintechNewsSGAsia has a three-year window to seize AI leadership through digital sovereignty - The Business Times
• Asia has a critical three-year window to establish global AI leadership by prioritizing digital sovereignty and rethinking digital borders. • To remain competitive, the region must integrate trust and security directly into the architecture of AI systems.
Read original · businesstimes.com.sgStocks dip on new Iran attacks as oil prices rise
• European stocks and U.S. futures declined on Wednesday following a rise in oil prices for the third consecutive session. • The market volatility was triggered by renewed hostilities in the Gulf after peace talks between the U.S. and Iran stalled.
Read original · reuters.com