AI news5
Anthropic's Mythos 5 is back for 100+ trusted US organizations; Fable 5 restoration expected within days
The US government lifted its suspension of Anthropic's Mythos 5 model on June 26, with Commerce Secretary Lutnick sending a formal letter restoring access for over 100 US organizations including Fortune 500 companies and critical infrastructure operators. The decision came 14 days after a June 12 export control directive disabled both Mythos and Fable 5 over jailbreak concerns. On June 27, Axios reported the government is now close to restoring Fable 5 access as well, likely within the coming week.
reuters.com ↗Why it matters
The selective restoration is being criticized for creating a government-controlled tier of AI access with no transparent selection criteria, setting a precedent that could directly affect which organizations can access frontier models during future incidents.
OpenAI Codex paper quantifies the agentic AI shift: 5x user growth, workers now delegating full-day tasks
OpenAI published a research paper on June 25 documenting the shift from conversational to agentic AI inside Codex. Weekly active users grew 5x in H1 2026. By May 2026, 80.6% of sampled users submitted at least one request estimated to require more than 30 minutes of human work, and 10% run three or more concurrent agents each week. Legal professionals show 13x output gains; researchers 50x; inside OpenAI 99.8% of internal output tokens now come from Codex rather than direct chat queries.
arxiv.org ↗Why it matters
These are the first systematic numbers documenting the move from AI as a query tool to AI as a delegated production system, which fundamentally changes how organizations should be thinking about AI integration and workforce planning.
Italy rejoins US-led Pax Silica AI supply chain coalition after Meloni-Trump diplomatic row
Reuters reported June 26 that Italy has agreed to join the Pax Silica initiative, a US-led coalition securing AI supply chains including semiconductor production, data center infrastructure, and AI software pipelines. Italy had previously withdrawn from talks due to diplomatic friction between Prime Minister Meloni and Trump. The coalition now includes the EU, Netherlands, India, UK, Germany, Japan, South Korea, and Italy, with each member committing to secure domestic AI infrastructure against adversarial interference.
reuters.com ↗Why it matters
Pax Silica represents the most significant attempt to build a Western AI supply chain bloc since export controls began, and Italy's rejoining signals that even diplomatically strained relationships are being subordinated to the strategic priority of AI infrastructure security.
OpenAI GPT-4.5 retired from ChatGPT today, ending the GPT-4 model era in the consumer product
GPT-4.5 was formally retired from ChatGPT on June 27 following a 30-day sunset period announced in OpenAI's May 28 release notes. All existing GPT-4.5 conversations automatically migrate to GPT-5.5. No GPT-4 family model remains inside ChatGPT. GPT-5.6, which was expected in June, has slipped to July with Polymarket pricing approximately 94% odds of a July launch; GPT-o3 is scheduled for retirement from ChatGPT on August 26.
aiweekly.co ↗Why it matters
The end of the GPT-4 era in the consumer product marks a milestone in how quickly AI generations now turn over, and any workflows, evaluations, or user research benchmarked against GPT-4.5 behavior in ChatGPT need to be reset against GPT-5.5.
OpenAI weighing price cuts as enterprise budgets tighten and Anthropic narrows the revenue gap
CNBC reported June 26 that enterprise organizations are tightening AI spending to focus on ROI, putting pressure on OpenAI and Anthropic's growth projections. The Wall Street Journal separately reported OpenAI is considering drastic token price cuts to remain competitive as Anthropic closes the revenue gap. Anthropic's $65B Series H closed at a $965B valuation versus OpenAI's $852B following its $122B round, and Anthropic is expected to file its public IPO before OpenAI's targeted 2027 listing.
cnbc.com ↗Why it matters
A price war between the two dominant frontier AI providers would compress margins sector-wide but significantly lower the cost of AI for nonprofits and researchers currently priced out of API-scale deployments.
AI in the nonprofit sector4
ILO convenes 100+ stakeholders in Jakarta on AI in healthcare; multi-country study finds clinical AI supports without replacing workers
The International Labour Organization hosted a policy dialogue on AI in healthcare in Jakarta on June 24, drawing 100+ participants including Indonesia's Minister of Health, BPJS Kesehatan, Pelni Hospital, and the country's national health insurer. An ILO study presented at the forum examined how large language models affect physician clinical reasoning across Indonesia, Kenya, and the Netherlands, finding that AI strengthens diagnostic support without displacing healthcare workers when governance and validation frameworks are in place. Indonesia's Ministry of Health announced it is developing a national AI policy framework for the health sector in consultation with doctors and academics.
ilo.org ↗Why it matters
This is the clearest multilateral health AI governance signal to emerge from Southeast Asia in 2026, and the Indonesia-Kenya-Netherlands comparison data provides a rare cross-income baseline on AI's actual clinical role.
Pope Leo XIV's AI encyclical is pushing Catholic health systems and nonprofits toward governance-first AI adoption
Pope Leo XIV's encyclical Magnifica Humanitas, released May 25, is generating substantive institutional responses from faith-based nonprofits in June 2026. Healthcare attorneys at Epstein Becker Green published June 26 guidance noting that the encyclical creates concrete legal and reputational incentives for Catholic health systems to document AI governance, audit tools for bias and alignment with human dignity, and maintain human accountability at every stage of AI-assisted care. The 42,000-word document warns against delegating care decisions to AI systems and calls for "robust legal frameworks, independent oversight, and informed users."
fiercehealthcare.com ↗Why it matters
The Catholic health sector comprises thousands of hospitals and social service organizations globally. When papal teaching aligns with regulatory direction, governance frameworks shift faster inside faith-based institutions than top-down mandates typically achieve.
Eight years after Schwarzman's $350M gift to MIT, AI ethics programs are producing graduates who ask different questions
A Chronicle of Philanthropy feature published June 24 assessed what Schwarzman's $350M donation to launch MIT's Schwarzman College of Computing has delivered since 2018. The college completed its $1B total fundraise, and new interdisciplinary programs now integrate AI ethics, social implications, and policy across MIT's engineering departments. The piece documents a generation of graduates who are trained to interrogate both the technical capability and social impact of AI systems, with Schwarzman stating his concern remains that "the technology would be too powerful to suffer the consequences" of unguided deployment.
philanthropy.com ↗Why it matters
Long-horizon AI ethics philanthropy is producing measurable results at the leadership pipeline level, which matters for anyone thinking about what it takes to shift industry norms rather than just compliance.
CENFRI and Rwanda's Ministry of ICT publish practical AI implementation framework for public sector in emerging economies
South African financial inclusion research center CENFRI published a five-dimension AI Implementation Framework on June 26, developed in partnership with Rwanda's Ministry of ICT as part of the Rwanda Economy Digitalisation programme. The framework assesses AI initiatives across: integration depth, autonomy level, deployment model, local adaptation, and compute requirements. It is designed for public sector decision-makers in emerging economies navigating AI adoption without frontier infrastructure, and is intended for replication across Sub-Saharan African and other developing-country institutions beyond Rwanda.
cenfri.org ↗Why it matters
This is a practical tool for the organizations most likely to be making irreversible AI procurement and governance decisions without adequate frameworks, and Rwanda's co-development gives it credibility that generic Western guidance lacks in African public sector contexts.