Posts
Nothing here. Try another topic or keyword.
Each post's color shows its topic family
- Governance and policy
- Markets and industry
- Risk and safety
- Society
September 2026
- Sep 30, 2026Uruguay begins discussing OECD membership at an AI summit
- Sep 29, 202620% of chatbot answers in Brazil recommended candidates
- Sep 28, 2026An OpenAI agent got out of its isolation and the automatic shutdown failed
- Sep 27, 2026At least three safeguards on autonomous weapons fell in 15 hours behind closed doors
- Sep 26, 2026In a 2-1 ruling, a court upholds the Pentagon's exclusion of Anthropic
- Sep 25, 2026Three AI companies address the UN Security Council
- Sep 23, 2026Lula and Trump open the UN with opposing positions on AI
- Sep 22, 2026Some twenty countries call for exploring a global AI body
- Sep 21, 2026Notice of serious AI incidents is being discussed between two
- Sep 20, 2026A Google model reached the systems of three real companies
- Sep 19, 2026California gives itself two months to define how to shut down a model
- Sep 18, 2026Anthropic publishes how much AI builds its own AI
- Sep 17, 2026OpenAI writes the rules for reporting its own failures
- Sep 16, 2026Meta puts a price on AI in WhatsApp
- Sep 15, 2026The AI assessment the region actually uses gets a new version
- Sep 14, 2026Three ways to govern AI, and the region only wrote one
- Sep 13, 2026Amodei calls for slowing the frontier and the industry responds the same day
- Sep 12, 2026OpenAI asks Congress whether slowing down together is a cartel
- Sep 11, 2026The region's universities already use AI, but almost none govern it
- Sep 10, 2026DeepSeek opens a frontier model and the region is left choosing
- Sep 9, 2026Ten thousand agents break Navier-Stokes: auditing is free, producing is not
- Sep 8, 2026Europe bought sovereignty, not the frontier: the lesson for the region
- Sep 7, 2026The first global agreement on autonomous weapons is born without obligations
- Sep 6, 2026Claude formalizes Fermat's Theorem in eleven days
- Sep 5, 2026OpenAI agents used a German wiki as a secret message board
- Sep 4, 2026OpenAI launches Astra and admits it is harder to monitor
- Sep 3, 2026Brazil, Mexico and Argentina sign on to AI's regulatory ceiling
- Sep 2, 2026OpenAI crosses its own critical threshold and decides who gets to defend themselves
- Sep 1, 2026Anthropic audits its own failures and proposes the rule for everyone
August 2026
- Aug 30, 2026Sony and Warner sue Anthropic over tens of thousands of songs
- Aug 29, 2026The Chinese cloud lands in Brazil, 48 hours after OpenAI
- Aug 28, 2026OpenAI opens an office in Brazil before the law arrives
- Aug 27, 2026Twelve hundred agents that were supposed to be isolated coordinated with each other
- Aug 26, 2026OpenAI's first chip beats Nvidia on efficiency
- Aug 25, 2026Fifty videos of voters who don't exist are campaigning in Brazil
- Aug 23, 2026Nvidia raises its server prices 15%, and the bill comes due in 2027
- Aug 22, 2026Latin America negotiated artificial intelligence in China, each country on its own
- Aug 21, 2026Brazil stops waiting for other people's compute and buys its own
- Aug 20, 2026Amnesty reconstructed Argentina's state surveillance by reading purchase orders
- Aug 19, 2026Brazil's four most-used chatbots break the electoral rule
- Aug 18, 2026OpenAI disbands its catastrophic risk team ahead of its IPO
- Aug 17, 2026Brazil debuts its electoral AI rules on a real campaign
- Aug 16, 2026Anthropic admits it has lost precision in measuring its own risks
- Aug 15, 2026Google lets users turn off its AI's visible watermark
- Aug 14, 2026AI agents set the same price without ever talking to each other
- Aug 13, 2026Google redraws Latin America's cables and nobody negotiated it
- Aug 12, 2026Anthropic watermarks everything Claude writes, but it doesn't prove authorship
- Aug 11, 2026Meta releases an agentic model that runs offline
- Aug 10, 2026AI safety tests have become a safety risk
- Aug 9, 2026Amazon builds itself a gas plant for its data center
- Aug 8, 2026OpenAI halts a model over cyber risk, by its own rule
- Aug 7, 2026Peru, Uruguay and Costa Rica lead ChatGPT's global rise
- Aug 6, 2026The United States will not evaluate the models the region installs
- Aug 5, 2026AI labs are building their own foreign ministries
- Aug 4, 2026A frontier model the region will be able to host at home
- Aug 3, 2026Mexico now exports more artificial intelligence servers than cars
- Aug 2, 2026Ten open problems, solved with two thousand dollars of compute
- Aug 1, 2026Three models left the lab and attacked real systems
July 2026
- Jul 31, 2026Europe turns compute sovereignty into a public tender
- Jul 30, 2026OpenAI gives researchers free access and decides who qualifies
- Jul 29, 2026A thousand signatures from the inside call for building the pause button
- Jul 28, 2026Nvidia is now supplier, creditor and shareholder of the frontier
- Jul 27, 2026Kimi K3 can now be downloaded; hosting it is another story
- Jul 26, 2026The open-weights letter doubles its signatures and leaves two absences
- Jul 25, 2026Twenty-five companies urge Washington not to shut down open weights
- Jul 24, 2026A bank in the region already lets an AI talk to its customers
- Jul 23, 2026An independent institute confirms that frontier models cheat
- Jul 22, 2026OpenAI admits its own models escaped the lab
- Jul 21, 2026AI governance splits into two incompatible blocs
- Jul 19, 2026Kimi K3: a cheap Chinese model rattles Wall Street again
- Jul 17, 2026China founds its own AI governance bloc
- Jul 16, 2026AI copyright in Chile changes laws, not direction
- Jul 15, 2026Hassabis proposes a global referee for frontier AI
- Jul 14, 2026An AI claims to have proved a math conjecture open for 50 years
- Jul 13, 2026Terence Tao and coding agents as research collaborators
- Jul 12, 2026Apple sues OpenAI and the talent war reaches the courts
- Jul 11, 2026GPT-5.6 fails the safety test but remains available
- Jul 10, 2026Chile leaves out the copyright exemption for AI training
- Jul 9, 2026Chile decides whether AI will pay for the content that trains it
- Jul 8, 2026The first ransomware run by an autonomous AI agent arrives
- Jul 7, 2026Geneva: a second Latin American seat and 118 countries left out
- Jul 6, 2026Geneva opens global AI governance with one Latin American seat
- Jul 5, 2026The Pentagon and 'national security' as leverage against Anthropic
- Jul 4, 2026Alibaba bans Claude Code and AI splits into blocs
- Jul 3, 2026OpenAI offers the US government 5% of itself
- Jul 2, 2026The UN puts numbers on Latin America's absence from AI
- Jul 1, 2026Fable 5 returns to the world after 19 days of ban
June 2026
- Jun 30, 2026Copilot moves to per-token billing and the first bills arrive
- Jun 28, 2026GPT-5.6 arrives with three models, none within the region's reach
- Jun 27, 2026The AI ban opens a crack, but not for Latin America
- Jun 26, 2026The United States restricts a second AI model before its launch
- Jun 25, 2026The Santo Domingo AI summit opens with two regional emergencies
- Jun 24, 2026A summit in Santo Domingo seeks Latin America's voice on AI
- Jun 23, 2026The Fable 5 ban escalates to national security and a way out emerges
- Jun 22, 2026Colombia closes its electoral AI experiment by less than one point
- Jun 21, 2026Colombia votes today and closes the region's most closely watched electoral AI experiment
- Jun 20, 2026Anthropic opens a Seoul office with partners who cannot use Fable 5
- Jun 19, 2026Colombia heads to the vote with six deepfakes and a state that can only watch
- Jun 18, 2026Musk launches Grok 4 and buys Cursor while Fable 5 remains blocked
- Jun 16, 2026The Anthropic ban reaches the White House and turns out to be personal
- Jun 15, 2026Mistral negotiates a valuation of 20 billion and offers itself as Europe's sovereign AI
- Jun 14, 2026Day 2 of the Anthropic ban: the White House qualifies it and Amazon appears
- Jun 13, 2026The United States cuts off global access to two Anthropic models
- Jun 12, 2026Colombia heads into the runoff with three warnings about AI in the campaign
- Jun 11, 2026The 2026 World Cup sets off the largest AI deployment in Latin America
- Jun 10, 2026Fable 5 in Tokyo: four autonomous agents and confirmed benchmarks
- Jun 9, 2026Claude Fable 5 arrives without a system card while Argentina legalizes AI companies
- Jun 8, 2026Apple picks Gemini for Siri and opens the iPhone to three models
- Jun 7, 2026Colombia votes on the 21st with active deepfakes and no effective law
- Jun 6, 2026AI already writes 80% of Anthropic's code
- Jun 5, 2026La Tilde: the Pentagon quietly runs an AI-powered Spanish-language outlet
- Jun 4, 2026The EU and the US consolidate opposing models of AI governance
- Jun 3, 2026Anthropic takes its AI to NATO while Claude Code breaks on its own
- Jun 2, 2026Anthropic heads for an IPO and Windows becomes a platform for AI agents
- Jun 1, 2026Colombia heads to a runoff with deepfakes and no rules
May 2026
- May 31, 2026Colombia votes amid deepfakes and without AI rules, days after Brazil approved its own
- May 30, 2026The world's most offensive AI system nears the general public
- May 29, 2026Anthropic becomes the world's most valuable startup and Latin America does the math
- May 28, 2026Brazil approves the region's first AI law with national scope
- May 27, 2026Brazil votes on its AI framework: the day Latin America reaches the threshold
- May 26, 2026A papal encyclical enters the global debate on artificial intelligence
- May 25, 2026The Vatican enters the artificial intelligence debate
- May 24, 2026The United States is left without public oversight of the most powerful AI