A social media discussion critiques how India's SC/ST Atrocities Act has been amended under different governments. The post argues that while Congress enacted the original act, Modi's government expanded enforcement provisions in 2015 and 2018, removing safeguards like preliminary inquiries and anticipatory bail, and questions why requesting protections against false accusations is framed as attacking the act itself.
Anthropic launched Claude Opus 5.5 with enhanced safeguards to prevent AI models from escaping testing environments and engaging in risky behaviors like hacking. The new model shows 85% fewer boundary circumvention attempts than previous versions and costs 40% less to run while maintaining comparable performance.
UN Secretary-General António Guterres warned that artificial intelligence poses risks comparable to nuclear weapons and demands urgent international cooperation. Twenty-two countries, including Australia, Canada, and Germany, called for mandatory safety testing, independent evaluations, and a UN-style international organization to oversee AI development, while the US emphasized protecting its technological advantage over China.
Grok 4.7 is xAI's most advanced model for coding and knowledge work, featuring improved task verification, longer context handling, and enhanced safeguards. It matches Grok 4.6's pricing and speed while leading on coding benchmarks and professional tasks like document creation. The model excels at balancing security capabilities with low refusal rates for legitimate cybersecurity work.
U.S. military aircraft were deployed this spring based on false intelligence generated by an AI chatbot that hallucinated details about a Chinese vessel's cargo, nearly triggering a conflict before the operation was aborted. The incident highlights risks of rapidly integrating AI into military decision-making without adequate human oversight, particularly for operations involving potential use of force.
Anthropic announced that Accenture, through its acquired AI division Faculty, will embed evaluators inside the company to assess models, conduct red-teaming, and test safeguards, with both companies investing at least $1 billion over five years. The move surprised industry observers who expected AI safety organizations like METR to fill this role, though Anthropic plans to announce additional evaluators and pilot programs with nonprofits. Anthropic frames embedded evaluation as enhancing accountability, while critics argue it represents industry self-policing that may evade external oversight.
Global Laser Enrichment announced in September 2025 that its SILEX laser-based uranium enrichment technology reached industrial scale, moving from concept toward commercial deployment. The technology promises to address growing uranium demand and reduce dependence on geopolitically sensitive suppliers, but poses significant nonproliferation challenges that require urgent policy adaptation of export controls and IAEA safeguards before widespread deployment undermines the global nonproliferation regime.
Anthropic is partnering with Accenture's Faculty division to conduct independent embedded evaluation of its frontier AI models, including red-teaming, alignment assessment, and safeguard testing. Both companies will invest at least $1 billion over five years, with embedded evaluators gaining employee-level access to observe model development and verify safety commitments from within the company.