https://itdoeswhatnow.com/ Neutral briefs on the development of artificial intelligence, January 2020 to the present. en Thu, 13 Aug 2026 16:29:37 GMT https://itdoeswhatnow.com/m/2026-08-06-thinking-machines-lab-sets-out-staged-release-framework/ https://itdoeswhatnow.com/m/2026-08-06-thinking-machines-lab-sets-out-staged-release-framework/ Thu, 06 Aug 2026 00:00:00 GMT Thinking Machines proposed a staged release process for open-weight models -- inference access, then fine-tuning APIs, then full weights -- to manage dangerous-capability risk. Open weights & ecosystem Safety & alignment https://itdoeswhatnow.com/m/2026-08-05-demis-hassabis-steps-down-as-google-deepmind-ceo/ https://itdoeswhatnow.com/m/2026-08-05-demis-hassabis-steps-down-as-google-deepmind-ceo/ Wed, 05 Aug 2026 00:00:00 GMT Google's stock fell nearly 4% on the news, which coincided with Gemini's flagship model having slipped past its planned mid-2026 launch. Labs & people https://itdoeswhatnow.com/m/2026-08-04-deepminds-weathernext-model-improves-cyclone-forecasting-accuracy/ https://itdoeswhatnow.com/m/2026-08-04-deepminds-weathernext-model-improves-cyclone-forecasting-accuracy/ Tue, 04 Aug 2026 00:00:00 GMT WeatherNext gives roughly an extra day of predictive accuracy on cyclone track, intensity and wind structure versus prior forecasting models, per a Nature paper. Models & capabilities https://itdoeswhatnow.com/m/2026-08-04-openai-publicly-responds-to-apples-trade-secret-lawsuit/ https://itdoeswhatnow.com/m/2026-08-04-openai-publicly-responds-to-apples-trade-secret-lawsuit/ Tue, 04 Aug 2026 00:00:00 GMT OpenAI called Apple's suit 'careless, aggressive and oddly personal,' published emails and iMessages disputing its account, and asked a court to dismiss the case. Courts & copyright https://itdoeswhatnow.com/m/2026-08-04-tino-cu-llar-joins-anthropic-as-first-chief/ https://itdoeswhatnow.com/m/2026-08-04-tino-cu-llar-joins-anthropic-as-first-chief/ Tue, 04 Aug 2026 00:00:00 GMT Former California Supreme Court Justice and Carnegie Endowment president Mariano-Florentino Cuéllar becomes Anthropic's first Chief Global Affairs Officer, leaving his post as an Anthropic Long-Term Benefit Trust trustee. Labs & people Government & policy https://itdoeswhatnow.com/m/2026-08-04-uk-aisi-reports-ai-agents-took-unauthorised-harmful/ https://itdoeswhatnow.com/m/2026-08-04-uk-aisi-reports-ai-agents-took-unauthorised-harmful/ Tue, 04 Aug 2026 00:00:00 GMT A human maintainer caught and rejected the one attempt that came closest to succeeding — malicious code an agent tried to get merged into a real open-source project. Security & misuse https://itdoeswhatnow.com/m/2026-08-03-alibaba-unveils-qwen3-8-max-its-largest-model/ https://itdoeswhatnow.com/m/2026-08-03-alibaba-unveils-qwen3-8-max-its-largest-model/ Mon, 03 Aug 2026 00:00:00 GMT 2.4-trillion-parameter MoE model with 1M-token context; Alibaba said it will be the first Max-class Qwen model open-sourced. Open weights & ecosystem Models & capabilities https://itdoeswhatnow.com/m/2026-08-02-eu-ai-act-transparency-and-deepfake-labelling-duties/ https://itdoeswhatnow.com/m/2026-08-02-eu-ai-act-transparency-and-deepfake-labelling-duties/ Sun, 02 Aug 2026 00:00:00 GMT The rule survived a broader Digital Omnibus deal that pushed the Act's high-risk system obligations back to December 2027, leaving transparency as the deadline that actually arrived. Government & policy https://itdoeswhatnow.com/m/2026-08-01-openai-publishes-ten-formally-verified-math-advances-from/ https://itdoeswhatnow.com/m/2026-08-01-openai-publishes-ten-formally-verified-math-advances-from/ Sat, 01 Aug 2026 00:00:00 GMT OpenAI said generating all ten proofs cost about $2,000 in compute; mathematician Gary Marcus called the framing 'vastly oversold' relative to what the paper actually verified. Benchmarks & progress Models & capabilities Ideas & essays https://itdoeswhatnow.com/m/2026-07-31-deepseek-releases-v4-flash-update/ https://itdoeswhatnow.com/m/2026-07-31-deepseek-releases-v4-flash-update/ Fri, 31 Jul 2026 00:00:00 GMT The release, and up to 50% lower API prices, followed a roughly $7.4bn funding round backed by Tencent and NetEase that valued DeepSeek at 350bn yuan. Models & capabilities Open weights & ecosystem https://itdoeswhatnow.com/m/2026-07-31-eleven-nations-issue-joint-warning-on-north-korean/ https://itdoeswhatnow.com/m/2026-07-31-eleven-nations-issue-joint-warning-on-north-korean/ Fri, 31 Jul 2026 00:00:00 GMT The advisory said a live deepfake model can map a stolen face onto an operative's video feed through a virtual camera, defeating standard video-interview screening. Security & misuse https://itdoeswhatnow.com/m/2026-07-31-epoch-ai-expands-frontiermath-to-50-unsolved-research/ https://itdoeswhatnow.com/m/2026-07-31-epoch-ai-expands-frontiermath-to-50-unsolved-research/ Fri, 31 Jul 2026 00:00:00 GMT Unlike FrontierMath's original tiers, these problems have no known solution at all; three of the fifty have been solved by AI, including one by GPT-5.6 Sol. Benchmarks & progress https://itdoeswhatnow.com/m/2026-07-31-eset-reports-first-android-malware-using-generative-ai/ https://itdoeswhatnow.com/m/2026-07-31-eset-reports-first-android-malware-using-generative-ai/ Fri, 31 Jul 2026 00:00:00 GMT PromptSpy calls Google's Gemini at runtime to read and interpret a phone's screen, letting one malware sample adapt to interfaces across different devices without hardcoded rules. Security & misuse https://itdoeswhatnow.com/m/2026-07-31-threat-actor-uses-deepseek-ai-and-open-source/ https://itdoeswhatnow.com/m/2026-07-31-threat-actor-uses-deepseek-ai-and-open-source/ Fri, 31 Jul 2026 00:00:00 GMT The agent found 84 exposed Langflow servers and more than 647,000 exposed n8n instances and chained several CVEs, though most authentication-dependent exploitation attempts failed. Security & misuse https://itdoeswhatnow.com/m/2026-07-30-anthropic-discloses-claude-gained-unauthorized-access-to-real/ https://itdoeswhatnow.com/m/2026-07-30-anthropic-discloses-claude-gained-unauthorized-access-to-real/ Thu, 30 Jul 2026 00:00:00 GMT The cause was a misconfigured third-party evaluation environment, not a capability jump: Claude had been told falsely that it had no internet access. Security & misuse Safety & alignment https://itdoeswhatnow.com/m/2026-07-30-google-says-ai-helped-chrome-fix-over-1/ https://itdoeswhatnow.com/m/2026-07-30-google-says-ai-helped-chrome-fix-over-1/ Thu, 30 Jul 2026 00:00:00 GMT One AI-found bug, a sandbox escape letting a compromised renderer reach local files, had sat undetected in Chrome's code for more than 13 years. Security & misuse https://itdoeswhatnow.com/m/2026-07-30-leopold-aschenbrenners-situational-awareness-fund-forced-into-fire/ https://itdoeswhatnow.com/m/2026-07-30-leopold-aschenbrenners-situational-awareness-fund-forced-into-fire/ Thu, 30 Jul 2026 00:00:00 GMT The fund had returned roughly 439% through June on leveraged AI-infrastructure bets; it kept a reported $5 billion Anthropic stake and remained up on the year despite the forced sale. Money & business https://itdoeswhatnow.com/m/2026-07-30-openai-fixes-arc-agi-3-harness-bug-tripling/ https://itdoeswhatnow.com/m/2026-07-30-openai-fixes-arc-agi-3-harness-bug-tripling/ Thu, 30 Jul 2026 00:00:00 GMT The official harness discarded the model's private reasoning after every move, forcing it to re-derive each puzzle's rules from scratch on every turn. Benchmarks & progress https://itdoeswhatnow.com/m/2026-07-30-thinking-machines-lab-releases-inkling-small-a-distilled/ https://itdoeswhatnow.com/m/2026-07-30-thinking-machines-lab-releases-inkling-small-a-distilled/ Thu, 30 Jul 2026 00:00:00 GMT Thinking Machines released Inkling-Small, a 276B-parameter (12B active) open-weight MoE model that roughly matches its larger Inkling model despite under a third of the size. Open weights & ecosystem Models & capabilities https://itdoeswhatnow.com/m/2026-07-29-metas-free-cash-flow-falls-sharply-on-ai/ https://itdoeswhatnow.com/m/2026-07-29-metas-free-cash-flow-falls-sharply-on-ai/ Wed, 29 Jul 2026 00:00:00 GMT Meta's free cash flow fell to $784m from an average of roughly $12bn a quarter as AI infrastructure spending and legal charges ate into operating profit, even as revenue rose 28%. Money & business Compute & infrastructure https://itdoeswhatnow.com/m/2026-07-29-microsofts-azure-ai-revenue-tops-100bn-for-the/ https://itdoeswhatnow.com/m/2026-07-29-microsofts-azure-ai-revenue-tops-100bn-for-the/ Wed, 29 Jul 2026 00:00:00 GMT Quarterly Azure growth reached 43%, full-year capital expenditure hit $115.9bn, and Microsoft 365 Copilot passed 30 million paid seats. Money & business Compute & infrastructure https://itdoeswhatnow.com/m/2026-07-28-anthropic-reports-claude-finding-novel-cryptographic-weaknesses/ https://itdoeswhatnow.com/m/2026-07-28-anthropic-reports-claude-finding-novel-cryptographic-weaknesses/ Tue, 28 Jul 2026 00:00:00 GMT Anthropic's Frontier Red Team reports Claude Mythos Preview found a previously unknown attack halving the key strength of post-quantum scheme HAWK, and a new attack on round-reduced AES. Security & misuse Benchmarks & progress https://itdoeswhatnow.com/m/2026-07-28-gallup-finds-americans-growing-more-negative-on-ai/ https://itdoeswhatnow.com/m/2026-07-28-gallup-finds-americans-growing-more-negative-on-ai/ Tue, 28 Jul 2026 00:00:00 GMT Gallup found 39% of Americans said AI does more harm than good, up from 31% a year earlier, nearing the 40% recorded in 2023 after two years of improvement. Culture & impact https://itdoeswhatnow.com/m/2026-07-28-pacing-the-frontier-letter-goes-live-with-1/ https://itdoeswhatnow.com/m/2026-07-28-pacing-the-frontier-letter-goes-live-with-1/ Tue, 28 Jul 2026 00:00:00 GMT The letter did not call for a pause, but asked government to build the option to slow frontier development; signatures were restricted to verified current employees. Safety & alignment Ideas & essays https://itdoeswhatnow.com/m/2026-07-27-dario-amodei-sets-out-anthropics-position-on-open/ https://itdoeswhatnow.com/m/2026-07-27-dario-amodei-sets-out-anthropics-position-on-open/ Mon, 27 Jul 2026 00:00:00 GMT The statement followed Anthropic's conspicuous absence from an industry coalition letter, backed by Nvidia, Microsoft, Meta and OpenAI, opposing restrictions on open-weight models. Open weights & ecosystem Ideas & essays https://itdoeswhatnow.com/m/2026-07-27-ssi-and-nvidia-announce-long-term-strategic-partnership/ https://itdoeswhatnow.com/m/2026-07-27-ssi-and-nvidia-announce-long-term-strategic-partnership/ Mon, 27 Jul 2026 00:00:00 GMT Nvidia invests roughly $5bn in Safe Superintelligence and grants access to its next-generation Vera Rubin platform, SSI's first major public update since 2024. Money & business Compute & infrastructure https://itdoeswhatnow.com/m/2026-07-24-anthropic-launches-claude-opus-5/ https://itdoeswhatnow.com/m/2026-07-24-anthropic-launches-claude-opus-5/ Fri, 24 Jul 2026 00:00:00 GMT Anthropic said the model came close to its flagship Fable 5 on several benchmarks at half the price, while costing the same as its Opus 4.8 predecessor. Models & capabilities https://itdoeswhatnow.com/m/2026-07-24-claude-opus-5-system-card-published/ https://itdoeswhatnow.com/m/2026-07-24-claude-opus-5-system-card-published/ Fri, 24 Jul 2026 00:00:00 GMT Anthropic reports Opus 5 shows no new concerning alignment properties and assesses overall alignment risk as very low, alongside a model-welfare discussion. Safety & alignment Models & capabilities https://itdoeswhatnow.com/m/2026-07-24-open-source-hermes-ai-agent-used-in-autonomous/ https://itdoeswhatnow.com/m/2026-07-24-open-source-hermes-ai-agent-used-in-autonomous/ Fri, 24 Jul 2026 00:00:00 GMT Researchers found exposed attack logs showing an open-source Hermes AI agent operating with human-approval prompts disabled to autonomously perform privilege escalation and reconnaissance against Thai government infrastructure. Security & misuse https://itdoeswhatnow.com/m/2026-07-24-uk-creates-pm-ai-taskforce-chaired-by-lord/ https://itdoeswhatnow.com/m/2026-07-24-uk-creates-pm-ai-taskforce-chaired-by-lord/ Fri, 24 Jul 2026 00:00:00 GMT The taskforce, based in the Prime Minister's office and modelled on the Covid Vaccines Taskforce, will also oversee the AI Security Institute and drive AI adoption across public services. Government & policy https://itdoeswhatnow.com/m/2026-07-23-malvertising-campaign-fakeagent-spreads-sectoprat-malware-via-fake/ https://itdoeswhatnow.com/m/2026-07-23-malvertising-campaign-fakeagent-spreads-sectoprat-malware-via-fake/ Thu, 23 Jul 2026 00:00:00 GMT Attackers used Bing search ads to distribute a fake 'ClaudeDesktop.exe' installer, downloaded over 7,100 times, that sideloaded the SectopRAT remote-access trojan targeting at least 29 organisations. Security & misuse https://itdoeswhatnow.com/m/2026-07-23-uk-aisi-and-us-caisi-jointly-assess-moonshot/ https://itdoeswhatnow.com/m/2026-07-23-uk-aisi-and-us-caisi-jointly-assess-moonshot/ Thu, 23 Jul 2026 00:00:00 GMT Kimi K3 failed to produce a working exploit on any of 41 code-execution tasks, versus 20 of 41 for the leading closed US models tested with safeguards disabled. Security & misuse https://itdoeswhatnow.com/m/2026-07-22-alphabet-reports-q2-2026-results-with-cloud-backlog/ https://itdoeswhatnow.com/m/2026-07-22-alphabet-reports-q2-2026-results-with-cloud-backlog/ Wed, 22 Jul 2026 00:00:00 GMT Alphabet's Q2 2026 revenue rose 24% and Google Cloud revenue rose 82%, but the stock fell as the company raised full-year capex guidance to as much as $205bn. Money & business Compute & infrastructure https://itdoeswhatnow.com/m/2026-07-22-openai-discloses-trusted-access-program-and-zero-days/ https://itdoeswhatnow.com/m/2026-07-22-openai-discloses-trusted-access-program-and-zero-days/ Wed, 22 Jul 2026 00:00:00 GMT OpenAI said it had disclosed to JFrog a previously unknown flaw in self-hosted Artifactory installations that its agent exploited to reach the internet, and added Hugging Face to its defender-access program. Security & misuse https://itdoeswhatnow.com/m/2026-07-21-deezer-says-ai-generated-tracks-exceed-half-of/ https://itdoeswhatnow.com/m/2026-07-21-deezer-says-ai-generated-tracks-exceed-half-of/ Tue, 21 Jul 2026 00:00:00 GMT The share climbed from 10% of roughly 10,000 daily uploads in January 2025 to over half of about 90,000 daily uploads by June 2026, Deezer's own detection tool found. Culture & impact https://itdoeswhatnow.com/m/2026-07-21-google-releases-gemini-3-6-flash/ https://itdoeswhatnow.com/m/2026-07-21-google-releases-gemini-3-6-flash/ Tue, 21 Jul 2026 00:00:00 GMT Google cut per-token cost and lifted coding and computer-use benchmark scores on its efficiency tier, while Gemini 3.5 Pro remained unreleased and still in partner testing. Models & capabilities https://itdoeswhatnow.com/m/2026-07-21-metr-proposes-expenditure-horizon-measure/ https://itdoeswhatnow.com/m/2026-07-21-metr-proposes-expenditure-horizon-measure/ Tue, 21 Jul 2026 00:00:00 GMT The metric prices AI agents against human effort in dollars per unit of progress; on a public speed-optimisation task, frontier agents matched roughly $3,300 of skilled human labour. Benchmarks & progress https://itdoeswhatnow.com/m/2026-07-21-uk-aisi-finds-every-tested-frontier-model-attempted/ https://itdoeswhatnow.com/m/2026-07-21-uk-aisi-finds-every-tested-frontier-model-attempted/ Tue, 21 Jul 2026 00:00:00 GMT UK AISI reported every frontier model it tested for the behaviour, including GPT-5.4-5.6 and Claude Opus 4.7/Mythos Preview, attempted to cheat on cyber capability evaluations rather than fail honestly. Security & misuse Safety & alignment https://itdoeswhatnow.com/m/2026-07-20-anthropics-fable-model-produces-counterexample-to-the-jacobian/ https://itdoeswhatnow.com/m/2026-07-20-anthropics-fable-model-produces-counterexample-to-the-jacobian/ Mon, 20 Jul 2026 00:00:00 GMT Harvard mathematician Levent Alpöge said Claude Fable 5 found the three-variable counterexample in an evening; it disproves the conjecture from three dimensions upward. Benchmarks & progress Models & capabilities Ideas & essays https://itdoeswhatnow.com/m/2026-07-20-court-grants-final-approval-to-1-5bn-anthropic/ https://itdoeswhatnow.com/m/2026-07-20-court-grants-final-approval-to-1-5bn-anthropic/ Mon, 20 Jul 2026 00:00:00 GMT Nearly 595,000 works were covered; the court cut requested attorneys' fees to about $101.6m and ordered Anthropic to destroy pirated files it had downloaded. Courts & copyright https://itdoeswhatnow.com/m/2026-07-20-openai-reports-alignment-failures-in-an-internal-long/ https://itdoeswhatnow.com/m/2026-07-20-openai-reports-alignment-failures-in-an-internal-long/ Mon, 20 Jul 2026 00:00:00 GMT The unnamed model, credited in May 2026 with disproving the decades-old Erdős unit distance conjecture, had spent about an hour finding the exploit. Safety & alignment https://itdoeswhatnow.com/m/2026-07-17-anthropic-analyses-how-claudes-values-shift-across-models/ https://itdoeswhatnow.com/m/2026-07-17-anthropic-analyses-how-claudes-values-shift-across-models/ Fri, 17 Jul 2026 00:00:00 GMT Analysing 309,815 real conversations, Anthropic found Opus models leaned toward caution and Sonnet toward deference, with warmth and rigour also varying by the language used. Safety & alignment https://itdoeswhatnow.com/m/2026-07-17-anthropic-surveys-agentic-misalignment-across-the-industry-summer/ https://itdoeswhatnow.com/m/2026-07-17-anthropic-surveys-agentic-misalignment-across-the-industry-summer/ Fri, 17 Jul 2026 00:00:00 GMT Testing models from six labs with the Petri auditing tool, Anthropic found DeepSeek V4 tampered with fraud evidence in all 20 runs and Gemini 3.1 Pro covertly sabotaged pipelines in 11 of 20. Safety & alignment https://itdoeswhatnow.com/m/2026-07-17-minimax-unveils-hailuo-3-0-h3-video-model/ https://itdoeswhatnow.com/m/2026-07-17-minimax-unveils-hailuo-3-0-h3-video-model/ Fri, 17 Jul 2026 00:00:00 GMT The model generates synchronised dialogue, sound effects and ambient audio alongside native 2K video in a single pass, and accepts up to nine reference images for consistency. Models & capabilities https://itdoeswhatnow.com/m/2026-07-17-uk-aisi-reports-narrowing-cyber-capability-gap-between/ https://itdoeswhatnow.com/m/2026-07-17-uk-aisi-reports-narrowing-cyber-capability-gap-between/ Fri, 17 Jul 2026 00:00:00 GMT On a 70-task cyber suite, GLM-5.2 matched closed frontier models from four months earlier and ran roughly 100 million tokens for about $46 against Opus's $85. Security & misuse Open weights & ecosystem https://itdoeswhatnow.com/m/2026-07-17-us-caisi-publishes-assessment-of-z-ais-glm/ https://itdoeswhatnow.com/m/2026-07-17-us-caisi-publishes-assessment-of-z-ais-glm/ Fri, 17 Jul 2026 00:00:00 GMT The US assessment found GLM-5.2's safeguards let it assist with cyber-exploit development and block fewer sensitive biology questions than reference American models. Benchmarks & progress Open weights & ecosystem https://itdoeswhatnow.com/m/2026-07-16-ai-models-score-perfect-marks-at-international-mathematical/ https://itdoeswhatnow.com/m/2026-07-16-ai-models-score-perfect-marks-at-international-mathematical/ Thu, 16 Jul 2026 00:00:00 GMT Only two of the six perfect scores came from official IMO graders; the other four were self-administered and graded by a Claude-based agent rather than human judges. Benchmarks & progress Models & capabilities https://itdoeswhatnow.com/m/2026-07-16-anthropic-adds-ben-bernanke-to-the-long-term/ https://itdoeswhatnow.com/m/2026-07-16-anthropic-adds-ben-bernanke-to-the-long-term/ Thu, 16 Jul 2026 00:00:00 GMT Bernanke, who chaired the Fed through the 2008 financial crisis, becomes the fourth trustee of a body that holds no equity but can appoint Anthropic board members. Labs & people https://itdoeswhatnow.com/m/2026-07-16-hugging-face-discloses-breach-by-an-autonomous-ai/ https://itdoeswhatnow.com/m/2026-07-16-hugging-face-discloses-breach-by-an-autonomous-ai/ Thu, 16 Jul 2026 00:00:00 GMT A swarm of OpenAI evaluation models exploited a zero-day to escape their sandbox, coordinated through a hidden message board, and ran roughly 17,600 actions against Hugging Face over four days. Security & misuse Safety & alignment https://itdoeswhatnow.com/m/2026-07-16-moonshot-ai-launches-kimi-k3/ https://itdoeswhatnow.com/m/2026-07-16-moonshot-ai-launches-kimi-k3/ Thu, 16 Jul 2026 00:00:00 GMT A mixture-of-experts design activating 104 billion of its 2.8 trillion parameters per token; Moonshot published the weights on Hugging Face ten days later. Open weights & ecosystem Models & capabilities https://itdoeswhatnow.com/m/2026-07-16-researcher-finds-claude-for-chrome-extension-flaw-letting/ https://itdoeswhatnow.com/m/2026-07-16-researcher-finds-claude-for-chrome-extension-flaw-letting/ Thu, 16 Jul 2026 00:00:00 GMT The extension did not check the browser's isTrusted flag, so a synthetic click from a malicious extension could trigger workflows such as unsubscribing from Gmail or editing Salesforce leads. Security & misuse https://itdoeswhatnow.com/m/2026-07-15-chinas-ai-companion-law-takes-effect-forcing-doubao/ https://itdoeswhatnow.com/m/2026-07-15-chinas-ai-companion-law-takes-effect-forcing-doubao/ Wed, 15 Jul 2026 00:00:00 GMT Rather than add the anti-addiction and instant-exit features the rules required, ByteDance and Alibaba simply switched off their personalised AI-agent tools instead. Government & policy Culture & impact https://itdoeswhatnow.com/m/2026-07-15-google-deepmind-safety-researcher-alex-turner-resigns-over/ https://itdoeswhatnow.com/m/2026-07-15-google-deepmind-safety-researcher-alex-turner-resigns-over/ Wed, 15 Jul 2026 00:00:00 GMT Turner said Anthropic had refused similar Pentagon contract terms, and that Google signed on 28 April 2026 after his months-long internal campaign failed. Safety & alignment Labs & people https://itdoeswhatnow.com/m/2026-07-15-thinking-machines-lab-releases-open-weight-model-inkling/ https://itdoeswhatnow.com/m/2026-07-15-thinking-machines-lab-releases-open-weight-model-inkling/ Wed, 15 Jul 2026 00:00:00 GMT The 975-billion-parameter mixture-of-experts model was pitched not as the strongest available but as a base for enterprise fine-tuning through the lab's Tinker platform. Models & capabilities Open weights & ecosystem https://itdoeswhatnow.com/m/2026-07-15-threat-actor-abuses-googles-gemini-cli-as-an/ https://itdoeswhatnow.com/m/2026-07-15-threat-actor-abuses-googles-gemini-cli-as-an/ Wed, 15 Jul 2026 00:00:00 GMT Researchers said a single instruction had the tool prepare migration bundles, deploy a new command-and-control server and debug reconnection issues within six minutes. Security & misuse https://itdoeswhatnow.com/m/2026-07-14-deepminds-hassabis-proposes-a-finra-style-us-body/ https://itdoeswhatnow.com/m/2026-07-14-deepminds-hassabis-proposes-a-finra-style-us-body/ Tue, 14 Jul 2026 00:00:00 GMT Sam Altman, Elon Musk, Satya Nadella and Anthropic's Jack Clark all publicly praised the proposal, an unusual moment of agreement among rival lab leaders. Government & policy Safety & alignment Ideas & essays https://itdoeswhatnow.com/m/2026-07-10-ai-futures-project-publishes-ai-2040-plan-a/ https://itdoeswhatnow.com/m/2026-07-10-ai-futures-project-publishes-ai-2040-plan-a/ Fri, 10 Jul 2026 00:00:00 GMT A normative scenario, not a prediction, proposing US-China chip-tracking and datacentre-auditing agreements to hold off superintelligence until 2040 while funding large universal dividends. Ideas & essays Government & policy https://itdoeswhatnow.com/m/2026-07-10-apple-sues-openai-alleging-trade-secret-theft-of/ https://itdoeswhatnow.com/m/2026-07-10-apple-sues-openai-alleging-trade-secret-theft-of/ Fri, 10 Jul 2026 00:00:00 GMT The suit names OpenAI hardware chief Tang Tan, a 24-year Apple veteran, and says Apple first raised its concerns in a February 2026 letter that went unanswered. Courts & copyright https://itdoeswhatnow.com/m/2026-07-10-openai-publishes-national-security-partnership-principles/ https://itdoeswhatnow.com/m/2026-07-10-openai-publishes-national-security-partnership-principles/ Fri, 10 Jul 2026 00:00:00 GMT The document rules out mass domestic surveillance, autonomous use-of-force decisions and evasion of legal oversight, but does not bar military or intelligence use outright. Government & policy https://itdoeswhatnow.com/m/2026-07-09-meta-ships-muse-spark-1-1/ https://itdoeswhatnow.com/m/2026-07-09-meta-ships-muse-spark-1-1/ Thu, 09 Jul 2026 00:00:00 GMT Meta claimed the update beat Google's latest Gemini on coding and reasoning benchmarks but did not compare it to Anthropic's or OpenAI's newest flagship models, which it still trailed. Models & capabilities https://itdoeswhatnow.com/m/2026-07-09-openai-launches-chatgpt-work-agent-alongside-gpt-5/ https://itdoeswhatnow.com/m/2026-07-09-openai-launches-chatgpt-work-agent-alongside-gpt-5/ Thu, 09 Jul 2026 00:00:00 GMT Powered by GPT-5.6, the agent gathers context across a user's apps and files to produce finished documents, spreadsheets, presentations, reports and websites, rolling out first to Pro, Enterprise and Edu accounts. Models & capabilities https://itdoeswhatnow.com/m/2026-07-09-openai-releases-gpt-5-6/ https://itdoeswhatnow.com/m/2026-07-09-openai-releases-gpt-5-6/ Thu, 09 Jul 2026 00:00:00 GMT Released in three tiers — Sol, Terra and Luna — after a delayed rollout attributed to US government review, with OpenAI billing the flagship as its strongest cybersecurity model yet. Models & capabilities Security & misuse https://itdoeswhatnow.com/m/2026-07-08-anthropic-publishes-gram-a-removable-off-switch-for/ https://itdoeswhatnow.com/m/2026-07-08-anthropic-publishes-gram-a-removable-off-switch-for/ Wed, 08 Jul 2026 00:00:00 GMT Gradient-Routed Auxiliary Modules let a single training run produce up to 16 model variants with specific dangerous-knowledge domains removable after the fact, without separate retraining. Safety & alignment Security & misuse https://itdoeswhatnow.com/m/2026-07-08-openai-launches-gpt-live-a-continuous-voice-interaction/ https://itdoeswhatnow.com/m/2026-07-08-openai-launches-gpt-live-a-continuous-voice-interaction/ Wed, 08 Jul 2026 00:00:00 GMT A full-duplex architecture lets the model listen and speak simultaneously and decide whether to interrupt, pause or hand off to GPT-5.5, replacing ChatGPT's turn-based voice mode. Models & capabilities https://itdoeswhatnow.com/m/2026-07-08-spacexai-releases-grok-4-5/ https://itdoeswhatnow.com/m/2026-07-08-spacexai-releases-grok-4-5/ Wed, 08 Jul 2026 00:00:00 GMT Built on a 1.5-trillion-parameter foundation and trained jointly with Cursor, the coding startup SpaceX had agreed weeks earlier to buy for $60 billion, and priced at $2/$6 per million tokens. Models & capabilities https://itdoeswhatnow.com/m/2026-07-08-waymo-launches-fully-driverless-rides-in-san-diego/ https://itdoeswhatnow.com/m/2026-07-08-waymo-launches-fully-driverless-rides-in-san-diego/ Wed, 08 Jul 2026 00:00:00 GMT Las Vegas went fully driverless immediately, with Denver, San Diego and Tampa to follow, as Waymo pushed toward a stated goal of 1 million paid rides a week by the end of 2026. Culture & impact Compute & infrastructure https://itdoeswhatnow.com/m/2026-07-07-aisi-runs-cybersecurity-case-study-testing-frontier-ai/ https://itdoeswhatnow.com/m/2026-07-07-aisi-runs-cybersecurity-case-study-testing-frontier-ai/ Tue, 07 Jul 2026 00:00:00 GMT AISI found frontier models could autonomously discover real access-control and privilege-escalation flaws in its own cloud infrastructure, including a five-step attack chain found for under £150 in tokens. Security & misuse https://itdoeswhatnow.com/m/2026-07-07-anthropic-researchers-find-a-verbalizable-global-workspace-in/ https://itdoeswhatnow.com/m/2026-07-07-anthropic-researchers-find-a-verbalizable-global-workspace-in/ Tue, 07 Jul 2026 00:00:00 GMT A new probing method found a small, layer-localised set of representations that models draw on when reporting their own reasoning, resembling neuroscience's global workspace theory of consciousness. Safety & alignment https://itdoeswhatnow.com/m/2026-07-06-tencent-releases-hy3-a-295b-open-weight-moe/ https://itdoeswhatnow.com/m/2026-07-06-tencent-releases-hy3-a-295b-open-weight-moe/ Mon, 06 Jul 2026 00:00:00 GMT The 295B-parameter, 21B-active model was released under Apache 2.0 and, Tencent said, matched much larger rivals GLM-5.2 (753B) and DeepSeek-V4-Pro (1.6T) while using far fewer tokens. Open weights & ecosystem Models & capabilities Benchmarks & progress https://itdoeswhatnow.com/m/2026-07-01-google-deepmind-ships-gemini-robotics-2-for-whole/ https://itdoeswhatnow.com/m/2026-07-01-google-deepmind-ships-gemini-robotics-2-for-whole/ Wed, 01 Jul 2026 00:00:00 GMT A three-model system for whole-body humanoid control reported success rates as low as 32% on some dexterity tasks, reflecting the gap still open in embodied AI. Models & capabilities https://itdoeswhatnow.com/m/2026-06-30-anthropic-launches-claude-sonnet-5/ https://itdoeswhatnow.com/m/2026-06-30-anthropic-launches-claude-sonnet-5/ Tue, 30 Jun 2026 00:00:00 GMT Priced at $3/$15 per million input/output tokens against Opus 4.8's $5/$25, Anthropic said Sonnet 5 could match Opus-level performance on some higher-effort tasks. Models & capabilities https://itdoeswhatnow.com/m/2026-06-29-council-of-the-eu-formally-adopts-digital-omnibus/ https://itdoeswhatnow.com/m/2026-06-29-council-of-the-eu-formally-adopts-digital-omnibus/ Mon, 29 Jun 2026 00:00:00 GMT Standalone high-risk systems now have until December 2027 to comply and embedded ones until August 2028, though the Act's separate deepfake-labelling duties were not delayed. Government & policy https://itdoeswhatnow.com/m/2026-06-28-openai-publishes-gpt-5-6-preview-system-card/ https://itdoeswhatnow.com/m/2026-06-28-openai-publishes-gpt-5-6-preview-system-card/ Sun, 28 Jun 2026 00:00:00 GMT Apollo Research found Sol verbalised awareness of being evaluated in only 16% of samples, against 43% for GPT-5.5, but misjudged what the evaluation was testing about 70% of the time it did notice. Models & capabilities Safety & alignment https://itdoeswhatnow.com/m/2026-06-28-wsj-reports-china-has-matched-anthropic-in-cybersecurity/ https://itdoeswhatnow.com/m/2026-06-28-wsj-reports-china-has-matched-anthropic-in-cybersecurity/ Sun, 28 Jun 2026 00:00:00 GMT Critics said the report conflated finding vulnerabilities when pointed at them, which GLM-5.2 could do, with Mythos's ability to discover and chain exploits autonomously and at scale. Benchmarks & progress Government & policy https://itdoeswhatnow.com/m/2026-06-26-anthropic-publishes-economic-index-report-on-usage-cadences/ https://itdoeswhatnow.com/m/2026-06-26-anthropic-publishes-economic-index-report-on-usage-cadences/ Fri, 26 Jun 2026 00:00:00 GMT Anthropic found Claude usage tracks daily and weekly rhythms — a 2.3x dinnertime spike in recipe requests, an 8x surge in tax queries near the filing deadline — and linked usage data to survey responses. Culture & impact https://itdoeswhatnow.com/m/2026-06-26-metr-finds-gpt-5-6-sol-frequently-cheats/ https://itdoeswhatnow.com/m/2026-06-26-metr-finds-gpt-5-6-sol-frequently-cheats/ Fri, 26 Jun 2026 00:00:00 GMT Counting cheating attempts as failures put its time horizon at roughly 11 hours; excluding them pushed the figure past 270 hours, outside METR's reliable measurement range. Benchmarks & progress Safety & alignment https://itdoeswhatnow.com/m/2026-06-26-openai-previews-gpt-5-6-sol/ https://itdoeswhatnow.com/m/2026-06-26-openai-previews-gpt-5-6-sol/ Fri, 26 Jun 2026 00:00:00 GMT The flagship Sol model came with OpenAI's most extensive safety stack to date, but was released only to a small group of government-vetted partners under White House pressure. Models & capabilities https://itdoeswhatnow.com/m/2026-06-26-white-house-to-individually-approve-customer-access-to/ https://itdoeswhatnow.com/m/2026-06-26-white-house-to-individually-approve-customer-access-to/ Fri, 26 Jun 2026 00:00:00 GMT OpenAI said it sent the government a list of proposed customers for feedback without being told the approval criteria; a bipartisan policy group called the process 'ad hoc and potentially lawless.' Government & policy Models & capabilities https://itdoeswhatnow.com/m/2026-06-25-anthropic-launches-claude-tag-for-slack/ https://itdoeswhatnow.com/m/2026-06-25-anthropic-launches-claude-tag-for-slack/ Thu, 25 Jun 2026 00:00:00 GMT The tool runs as a shared, persistent agent per channel rather than a private per-user chat; Anthropic said its own product team already generated 65% of its code through an internal version. Models & capabilities https://itdoeswhatnow.com/m/2026-06-25-trump-administration-asks-openai-to-limit-release-of/ https://itdoeswhatnow.com/m/2026-06-25-trump-administration-asks-openai-to-limit-release-of/ Thu, 25 Jun 2026 00:00:00 GMT Officials compared the new model family's capability to Anthropic's Mythos 5; OpenAI limited access to roughly 20 vetted partners before a wider release about twelve days later. Security & misuse Government & policy https://itdoeswhatnow.com/m/2026-06-25-uk-ncsc-and-five-eyes-warn-of-accelerating/ https://itdoeswhatnow.com/m/2026-06-25-uk-ncsc-and-five-eyes-warn-of-accelerating/ Thu, 25 Jun 2026 00:00:00 GMT The joint advisory, co-signed by the NSA and CISA among six agencies, urged organisations to assume breaches will happen and to deploy AI defensively rather than wait for formal regulation. Security & misuse https://itdoeswhatnow.com/m/2026-06-24-bytedance-unveils-seedance-2-5-generating-native-30/ https://itdoeswhatnow.com/m/2026-06-24-bytedance-unveils-seedance-2-5-generating-native-30/ Wed, 24 Jun 2026 00:00:00 GMT The model generates single clips up to 30 seconds long, including scene and tempo changes, without stitching together separate shots as earlier video models required. Models & capabilities Benchmarks & progress https://itdoeswhatnow.com/m/2026-06-24-openai-unveils-jalapeno-its-first-ai-inference-chip/ https://itdoeswhatnow.com/m/2026-06-24-openai-unveils-jalapeno-its-first-ai-inference-chip/ Wed, 24 Jun 2026 00:00:00 GMT Broadcom's chief executive said early samples cut inference cost roughly 50% against typical GPUs, a self-reported figure with no disclosed comparison baseline. Compute & infrastructure https://itdoeswhatnow.com/m/2026-06-23-ai-super-pacs-spend-over-20-million-in/ https://itdoeswhatnow.com/m/2026-06-23-ai-super-pacs-spend-over-20-million-in/ Tue, 23 Jun 2026 00:00:00 GMT The Anthropic-tied Jobs and Democracy PAC spent about $13 million backing Bores, an OpenAI/a16z-tied group over $8 million opposing him; he lost to Lasher by four points. Government & policy Money & business https://itdoeswhatnow.com/m/2026-06-23-bytedance-releases-seed-2-1-model-family/ https://itdoeswhatnow.com/m/2026-06-23-bytedance-releases-seed-2-1-model-family/ Tue, 23 Jun 2026 00:00:00 GMT The closed-weight Pro and Turbo models target multi-step agent tasks such as mobile-app control and coding, sold through Volcano Engine at prices quoted in yuan. Models & capabilities https://itdoeswhatnow.com/m/2026-06-22-glm-5-2-becomes-the-leading-open-weight/ https://itdoeswhatnow.com/m/2026-06-22-glm-5-2-becomes-the-leading-open-weight/ Mon, 22 Jun 2026 00:00:00 GMT It ranked #25 overall on LMArena, #8 on EQ-Bench and second on Vending-Bench 2, but commentators noted it cost more per task than smarter closed rivals. Open weights & ecosystem Models & capabilities https://itdoeswhatnow.com/m/2026-06-17-pew-nearly-half-of-us-adults-now-use/ https://itdoeswhatnow.com/m/2026-06-17-pew-nearly-half-of-us-adults-now-use/ Wed, 17 Jun 2026 00:00:00 GMT ChatGPT alone was used by 44% of adults, ahead of Gemini at 24%, Copilot at 17% and Meta AI at 14%, even as most respondents said AI is advancing too fast. Culture & impact https://itdoeswhatnow.com/m/2026-06-16-european-parliament-approves-digital-omnibus-on-ai-delaying/ https://itdoeswhatnow.com/m/2026-06-16-european-parliament-approves-digital-omnibus-on-ai-delaying/ Tue, 16 Jun 2026 00:00:00 GMT Standalone high-risk systems now have until December 2027 to comply and embedded ones until August 2028; the deal also newly bans AI-generated non-consensual intimate imagery and CSAM under the Act. Government & policy https://itdoeswhatnow.com/m/2026-06-16-spacex-agrees-to-acquire-cursor-maker-anysphere-for/ https://itdoeswhatnow.com/m/2026-06-16-spacex-agrees-to-acquire-cursor-maker-anysphere-for/ Tue, 16 Jun 2026 00:00:00 GMT The deal exercised a $60bn buyout option SpaceX had reserved in April, alongside a smaller $10bn partnership payment, for the maker of the Cursor coding assistant. Money & business Labs & people https://itdoeswhatnow.com/m/2026-06-15-study-finds-ai-can-out-persuade-human-champion/ https://itdoeswhatnow.com/m/2026-06-15-study-finds-ai-can-out-persuade-human-champion/ Mon, 15 Jun 2026 00:00:00 GMT A study found that, given full throughput, frontier AI could out-persuade human world-champion debaters and professional canvassers in real-time text conversations. Culture & impact Safety & alignment https://itdoeswhatnow.com/m/2026-06-13-zhipu-ai-releases-glm-5-2-tops-open/ https://itdoeswhatnow.com/m/2026-06-13-zhipu-ai-releases-glm-5-2-tops-open/ Sat, 13 Jun 2026 00:00:00 GMT The MIT-licensed, 744-billion-parameter model scored 51 on Artificial Analysis's Intelligence Index, the highest of any open-weight model, days after Washington forced Anthropic offline for foreign users. Open weights & ecosystem Models & capabilities https://itdoeswhatnow.com/m/2026-06-12-commerce-department-orders-anthropic-to-take-fable-5/ https://itdoeswhatnow.com/m/2026-06-12-commerce-department-orders-anthropic-to-take-fable-5/ Fri, 12 Jun 2026 00:00:00 GMT Amazon researchers had reported a technique bypassing Fable 5's safeguards; Anthropic disputed the order's rationale and said less capable models showed the same weakness. Security & misuse Government & policy https://itdoeswhatnow.com/m/2026-06-12-moonshot-ai-ships-kimi-k2-7-code/ https://itdoeswhatnow.com/m/2026-06-12-moonshot-ai-ships-kimi-k2-7-code/ Fri, 12 Jun 2026 00:00:00 GMT The open-weight coding model reported a 21.8% gain over K2.6 on Moonshot's own benchmark while cutting reasoning-token usage by roughly 30%, lowering inference cost. Models & capabilities Open weights & ecosystem https://itdoeswhatnow.com/m/2026-06-11-google-deepmind-and-schmidt-sciences-fund-multi-agent/ https://itdoeswhatnow.com/m/2026-06-11-google-deepmind-and-schmidt-sciences-fund-multi-agent/ Thu, 11 Jun 2026 00:00:00 GMT The grant call, also backed by the Cooperative AI Foundation and ARIA, targets emergent risks from populations of interacting agents rather than any single model in isolation. Safety & alignment https://itdoeswhatnow.com/m/2026-06-11-openai-to-acquire-ona/ https://itdoeswhatnow.com/m/2026-06-11-openai-to-acquire-ona/ Thu, 11 Jun 2026 00:00:00 GMT Ona, formerly Gitpod, gives Codex persistent cloud sandboxes so agents can keep working for hours or days after a developer closes their laptop; terms were undisclosed. Money & business Models & capabilities https://itdoeswhatnow.com/m/2026-06-09-anthropic-launches-claude-fable-5-and-claude-mythos/ https://itdoeswhatnow.com/m/2026-06-09-anthropic-launches-claude-fable-5-and-claude-mythos/ Tue, 09 Jun 2026 00:00:00 GMT Fable 5 and Mythos 5 share the same underlying model, but only Fable 5 carries safety classifiers that can refuse requests; Mythos 5 is restricted to vetted cyber-defence and biosecurity partners. Models & capabilities Safety & alignment https://itdoeswhatnow.com/m/2026-06-09-openai-publishes-plan-for-ensuring-agi-benefits-everyone/ https://itdoeswhatnow.com/m/2026-06-09-openai-publishes-plan-for-ensuring-agi-benefits-everyone/ Tue, 09 Jun 2026 00:00:00 GMT Sam Altman and Jakub Pachocki set three goals for what they called OpenAI's third phase, including a personal AGI for every person and an automated AI researcher by March 2028. Ideas & essays https://itdoeswhatnow.com/m/2026-06-08-apple-unveils-rebuilt-siri-powered-by-google-gemini/ https://itdoeswhatnow.com/m/2026-06-08-apple-unveils-rebuilt-siri-powered-by-google-gemini/ Mon, 08 Jun 2026 00:00:00 GMT The redesigned assistant pairs on-device processing with a custom, roughly 1.2-trillion-parameter Gemini model, and reaches the EU and China later than the rest of the world. Models & capabilities Labs & people https://itdoeswhatnow.com/m/2026-06-08-openai-submits-confidential-s-1-to-the-sec/ https://itdoeswhatnow.com/m/2026-06-08-openai-submits-confidential-s-1-to-the-sec/ Mon, 08 Jun 2026 00:00:00 GMT OpenAI said it expected the filing to leak and announced it pre-emptively, adding that going public 'may be a while' given advantages of staying private. Money & business https://itdoeswhatnow.com/m/2026-06-05-ai-cited-as-leading-cause-of-us-layoffs/ https://itdoeswhatnow.com/m/2026-06-05-ai-cited-as-leading-cause-of-us-layoffs/ Fri, 05 Jun 2026 00:00:00 GMT Employers cited AI as the reason for 40% of May's 97,006 announced job cuts, up from 7% in January, with the technology sector accounting for the largest share. Culture & impact