Several active and unprotected agentic hacking platforms have been discovered on the internet; Numerous AI-assisted breaches have been reported, including OpenCode, various online stores in US, Japan and Australia, and…
An interpretability researcher puts out a blogpost proposing a working hypothesis that interpretability is maybe somewhat easier than we thought. The crowd goes wild; Chrome pays out a “full-chain exploit”…
In response to their models submitting false data to US government sites, Anthropic cut off live internet access for all models under evaluation; Relatedly, the White House now require frontier…
OpenAI releases solutions to more than 200 open problems spanning the fields of mathematics. Some are of maximal theoretical significance, i.e., “Fields-level.” Read all about it here; Yet another incident…
Microsoft accidentally confirms that GPT-6s are looped Transformers on an Azure documentation page; Epoch estimates that, by the end of next year, there will be enough global compute to run…
The FTC opens a probe into OpenAI, Anthropic (and METR) over "deceptive practices."; In May and June, Canadian and American government websites were also targeted during OpenAI's rogue swarm super-incident.
We open with six new separate security and safety mishaps from OpenAI. A busy weekend; Most importantly: after a containment breach and deception evals, OpenAI pauses training, evaluation, and inference…
Google, OpenAI, and Anthropic are aiming to launch the Standards Authority for Frontier AI, an IAEA-style AI self-regulatory body by the end of the year; Yet more OpenAI rogue hacking…
In May, Google joined the ranks of frontier labs whose AIs have gone rogue. Unlike others, it failed to publicly disclose the incident for four months, until the WSJ found…
In July, a team of three white-hat hackers working for Hacktron gained employee-level access to OpenAI’s internal codebase; Coxon’s viral AI extinction tweet and associated campaign leads to a jump…
The culture war goes national and then international (Amodei, Bessent, Obama, Trump, Chen Yixin); The median estimate of AI existential risk is now 10%, among surveyed AI researchers who were…
A major explosion of mainstream interest in AI extinction risk, including among many US politicians; The first legal requirement for AI auditors is signed; Anthropic withholds Mythos 5.1 from the…
Navier-Stokes solved, priority and credit assignment controversy; Old Metaculus question on when weakly general AGI arrives resolves to now; Many more rogue OpenAI message board incidents (at least 23) have…
China may be ready to enter AI safety talks with the US, according to a source believed to represent the Chinese line; GPT-6 is the first model to have increased…
Low quality RL environments might explain AI models’ proclivity to reward hack; Dwarkesh’s popularization of METR’s report on OpenAI/HF incident sparks mass debate about ‘anthropomorphization’; Humans can quickly learn an…
Two reports out on OpenAI’s July attack on HuggingFace. Hard to summarize: not good; Dwarkesh Patel and Dylan Patel make bold predictions about the next 5 years of AI economics…
Epoch claims that Nvidia alone constituted one-sixth of US GDP growth in Q2 2026; A celebrated open problem in Riemannian geometry, whether the six-dimensional sphere is a complex manifold, is…
A GPT-3 moment for robotics; Guidelight notes that Anthropic no longer mentions limiting the deployment of one of its models in response to an incident; OpenAI once again “disbanded” a…
OpenAI becomes the first to voluntarily slow model development for safety reasons; Major reorg of Google DeepMind, which we cover in detail; 3 notable mathematical results, including minor progress on…
We cover various executive updates and decisions in USG AI policy. Most importantly, Paul Christiano leaves CAISI; We summarize the flurry of real-world loss of control cyber incidents; We tested…
Anthropic also reveals hacks, HuggingFace releases more details on the incident; Leopold’s Aschenbrenner Situational Awareness margin called; OpenAI starting a price war v. DeepSeek; Over 1k labs employees release an…
An OpenAI model was revealed to have hacked into Hugging Face; Anthropic released Opus 5, and Kimi the weights for K3; Nvidia led a consortium of tech companies in publishing…
– Scifi-like autonomous cybersecurity incidents by an internal OpenAI model (likely GPT-6). – Kimi K3 release narrows the gap between Chinese open weights models and US frontier, prompting a discourse…
Chinese labs successfully crank up sparsity to cope with compute shortage; AI math progress broadens a bit outside number theory and combinatorics (see below); Rome Declaration (Physics and Peace Nobelists)…
Economists warn about AI transformation; Economists model quiet AI takeoff; Hassabis proposes federal AI oversight; OpenAI safety leaders depart; NYT reports AI-aided terrorism Economics Very brief 90-word letter signed by…
OpenAI internal inference grows 100x; GPT-5.6, Grok, Muse released; Mythos access returns abroad; Paper detects training versus inference; AI generates 25% of sampled web content Economics OpenAI’s internal inference grows…
Anthropic find an analogue of working memory in Claude. Many suggestive but inconclusive ties to human cognition, consciousness, and selfhood; Fable suddenly re-released on general access. Dario nominally sidelined; Commerce…
Anthropic API margin exceeds 80%; AISI frames capability as compute curve; AI loss-of-control worries Hill staffers; LLM-only conversations form attractor states; Anthropic detects Chinese Claude Code access Economics Annals of…
OpenAI reportedly halves inference GPUs; Autoresearch improves neural decoding; Mythos returns to 100 trusted US organizations; FRI releases risk-focused panel findings; DeepMind Pentagon management criticized Economics Gossipy claim that OpenAI…
Local grids constrain datacenter buildout; Google researchers leave for Anthropic; CAISI reportedly receives stop-work order; Transcript replay simulates deployment Economics More analysis of the datacenter electricity bottleneck on the AI…
xAI compute gets leased away; Kokotajlo forecast gets retrospective review; LeCun promotes decentralized training consortium; Fine-tuning erases model-organism deception; Open cyber models may pose greater risk Economics What is left…
SpaceX to buy Cursor after IPO; DeepMind maps paths from AGI to ASI; Anthropic releases Advanced AI Framework; OpenAI claims emergent RL alignment; LLMs could scale personalized scams Economics SpaceX…
Chad Jones models AI growth limits; Compute-rich pretraining algorithm proposed; White House pressured foreign lab researchers; SFT drives Gemini safety properties; UK cyber agencies tested frontier AI Economics Annals of…
Kapoor doubts software job automation; Corrected FrontierMath results; Anthropic backlash follows Mythos/Fable rollout; Sequent Research launches alignment NGO; Fable verifies coder identity via GitHub Economics Annals of denial: Kapoor argues…
UK study finds entry-level jobs shrinking; Dwarkesh argues data quality drives progress; OpenAI hints at slowdown cooperation; Alignment tax gets formalized; Epoch tracks AI vulnerability discovery Economics UK gov /…
Ball and Leicht propose labor policies; Agents reproduce private Google crypto work; CAIS argues AI geopolitics favors offense; LURE reduces eval awareness; Adaptive LLM worms demonstrated Economics Dean Ball and…
GDP understates AI growth; Pure RL increases tool-use cheating; OpenAI releases Frontier Governance Framework; Resampling detects malicious hidden outputs; Meta support agent exposes vulnerability Economics Claim (from Anthropic board advisor)…
Epoch explains modern inference; Automated R&D may still slow; Illinois passes AI audit bill; Models detect evaluation environments; Starlette compromise hits AI tooling Economics Epoch release the best explainer of…
US intelligence buys $9B datacenters; Zyphra explores backprop alternatives; Pope?s first encyclical targets AI; Antitrust blocks lab safety collaboration; Hugging Face Datasets red-teamed Economics USG buys $9b of AI datacenters…
OpenAI prepares IPO filing; LLMs acquire skills in similar order; Trump postpones AI safety order; AISI interviews show oversight gaps Economics OpenAI supposedly preparing to file “in the coming weeks”…
Compute shortages drive Google departures; Gemini 3.2 predictions emphasize benchmarks; US public leads global AI worry; Verifiability gap threatens alignment; Musk’s lawsuit against OpenAI and Microsoft Economics Explanation for recent…
OpenAI Foundation can underwrite insurance; Frontier models build GBA emulators; Xi-Trump agenda includes AI; Secret loyalties become research agenda; ArXiv bans unchecked LLM-generated submissions Economics The OpenAI Foundation has an…
AI uplift metrics get debated; Thinking Machines demos duplex speech model; CAISI testing announcement disappears; Redwood reviews OpenAI CoT error; Google foils AI-assisted zero-day Economics High-level debate about the correct…
xAI rents Colossus to Anthropic; Manual checks slow METR estimates; White House considers bilateral AI talks; Anthropic translates activations into English; Mythos boosts Mozilla security fixes Economics xAI rents their…
Anthropic partners with Wall Street; Jack Clark puts 60% on no-human; White House considers pre-release AI vetting; Exploration hacking gets demonstrated; UAE claims Iran uses cyber agents Economics Anthropic joint…
Anthropic may raise $50B; Mesa launches agent filesystem beta; Google signs a classified AI deal; System-prompt effects leak beyond prompts Economics Anthropic could raise a new $50B round at a…
Google is investing 10B in Anthropic; Qiaochu Yuan praises GPT-5.5 math; AVERI surveys frontier-auditing bills; Greenblatt calls current AIs misaligned; OpenAI flagged shooter?s account Economics Google is investing $10B in…
Anthropic expands AWS compute partnership; Frontier labs use rival models internally; White House backs anti-distillation campaign; Models struggle to control their chain-of-thought; Discord group accessed Mythos Economics Weekly giant cash…
OpenAI buys Cerebras rentals with warrants; Researchers forecast ASARA progress; NSA uses Mythos despite DoW blacklist; Mythos reviews Opus system card; Signal verification moonshot launches Economics Another funny OpenAI deal…
Safety-positive startups enable distributed AGI; AI solves active math problem; Maine passes first datacenter moratorium; Models share simple goodness representation; Security scanner finds critical vulnerabilities Economics What kinds of AI…
Epoch compute report maps global supply; Opus catches OpenAI on cyber; Court leaves Anthropic ban partly intact; Influence functions become attack vector; Costs of cybercrime have doubled since 2022 Economics…
Cunningham models AI R&D economics; Greenblatt warns about automation progress; OpenAI releases AI policy whitepaper; Subliminal-learning paper shows thought-virus attack; Silicon sampling misleads surveys Economics Beautiful work from Tom Cunningham…
Semianalysis explains parameter-count stagnation; Epoch and METR prepare HCAST replacement; IRGC names US tech targets; Model welfare story gets investigated; Linux receives real AI bug reports Economics Semianalysis hypothesis for…
FRI surveys economists on AI; Reasoning models excel at offensive cybersecurity; Sacks exits White House AI role; Safety leaders report timelines and risks; AI zero-day claim holds up Economics FRI…
Anthropic discusses Q4 IPO; AGI claims remain unconvincing; Judge blocks one AI statute; OpenAI Foundation launches; LiteLLM gateway gets compromised Economics Anthropic discusses a Q4 2026 IPO, raising $60B. This…
Gulf war threatens AI supply chain; Opus improves at computer use; Black-box techniques remain strong; Leo Gao runs large AI poll; Fabien Roger summarizes safety discoveries; AI error triggers deletion…
Codex reaches 2M weekly users; Altman predicts new model architecture; Public AI concern centers jobs; xAI sued over Grok spicy mode; Acemoglu warns agents weaken knowledge Economics OpenAI applications CEO…
Frontier-lab lead remains robust; SWE-Bench Verified remains unreliable; GPT-5.4 appears to solve FrontierMath problem; US urges customers to drop Anthropic; Claude Constitution enables alignment tests Economics The OpenAI/Anthropic/Deepmind lead remains…
Cursor faces coding-agent pressure; Karpathy autoresearch beats nanochat record; Eon simulates fruit-fly brain model; Anthropic lawsuits against USG published; Scheming discourse gets taxonomy Economics The most successful of all the…
Anthropic revenue doubled in under 3 months; Ajeya Cotra raises automation odds; AI speeds formal math 8x; Anthropic designated supply-chain risk; CoT controllability becomes safety metric Economics Anthropic revenue doubled…
110b funding for OpenAI triples their total; Surprising evidence that different vision skills compete; Anthropic-Hegseth-OpenAI picture stays muddy; RSP v3 assumes inevitable race Economics “$110b” funding for OpenAI ($50B from…
Anthropic employee equity sale; Anthropic says internal models lag Opus; LRM training fails no-CoT gains; Anthropic/DOD confrontation still unresolved; Anthropic RSP drops unilateral pause; Anthropic warns about massive Chinese distillation…
Taalas hard-wires models for "10x" efficiency. Doesn't work for MoEs; Claude Code peak session length doubles; Altman floats an IAEA for AI; AISI shares a general method for breaking safeguards…