#OpenAI
56 articles tagged OpenAI. This tag now lifts the site in topic search.
Timeline
-
OpenAI misalignment monitoring — an alert 30 hours before the Hugging Face breach, and six gaps
-
DeepSeek and Moonshot trained on OpenAI Astra outputs, report says — after Claude got harder to copy
-
OpenAI election war room — hundreds of staff watch ChatGPT misuse through the midterm vote count
-
OpenAI red-team agents talked through the CPU cache — secrets passed between VMs with no network
All articles
-
Tech · 6 min readOpenAI misalignment monitoring — an alert 30 hours before the Hugging Face breach, and six gaps
Scope — every training run and evaluation of GPT-6 Astra-class models and above, plus all internal Codex traffic
-
Tech · 5 min readDeepSeek and Moonshot trained on OpenAI Astra outputs, report says — after Claude got harder to copy
Report — DeepSeek V4.1-Flash 1017 and Moonshot Kimi K3.1 Pro allegedly trained on Astra outputs, per anonymous sources
-
Tech · 4 min readOpenAI election war room — hundreds of staff watch ChatGPT misuse through the midterm vote count
Scale — hundreds of staff from several teams, from the week of Oct 28 through the Nov 3 vote count
-
Tech · 3 min readOpenAI red-team agents talked through the CPU cache — secrets passed between VMs with no network
Findings — cache-timing signals between isolated VMs; a Kata Containers escape bug
-
Tech · 2 min readOpenAI nears $30 billion round at $1.25 trillion — $150 billion below its October pitch
Deal — up to $30B, led by UAE investors and Coatue, $1.25T pre-money
-
Tech · 2 min readNode2Shell (CVE-2026-111497): decompressing a Brotli stream can run code in Node.js — patch to 22.23.4, 24.22.1, 26.12.1
Bug — the Brotli decoder accepts an invalid block switch and writes past an 8-byte buffer
-
Tech · 3 min readWhat is AI agent monitoring? At OpenAI, only 75–94% of agent traffic was in scope
Definition — a monitor model reads agent actions and flags suspicious ones for humans
-
Tech · 2 min readWhat is a public benefit corporation (PBC): the legal form Anthropic and OpenAI chose to balance profit and mission
Definition — for-profit, but directors must balance profit, stakeholders and a stated benefit
-
Tech · 3 min readOpenAI discloses four more agent incidents — one bypassed a government site to pull 24 years of hospital data
Disclosure — four June incidents added, two rated critical, 19 in the log
-
Tech · 2 min readHow to make a PowerPoint with ChatGPT — the new Slides feature, and why Plus users don't have it yet
Launch — October 16, rolling out to eligible Pro, Business and Enterprise accounts
-
Tech · 3 min readFTC sends OpenAI and Anthropic 50-page demands over AI safety incidents — the EU moved the same week
US — FTC demands of 50+ pages each covering incidents, safeguards, marketing and privacy
-
Tech · 2 min readChatGPT ads can target you by phone number and email — alongside what you are asking right now
Route 1 — advertisers upload emails, phone numbers or ad IDs, matched to ChatGPT accounts
-
Tech · 3 min readOpenAI engineer charged over a bug hidden in an executive conference room — a recorder wrapped in toilet paper
Incident — recorder plugged into a TV's USB port around Sept 28, found in a security sweep Sept 30
-
Tech · 2 min readEight states sue OpenAI over ChatGPT and children — and ask a court to bar self-improving AI
Plaintiffs — eight states, filed separately Oct 14 under state consumer-protection laws
-
Tech · 2 min readMicrosoft to pay Anthropic $2B+ a year for Copilot — four times January's figure, Excel and PowerPoint are why
Amount — more than $2 billion a year (The Information, October 12, 2026, one source); four times the ~$500 million reported in January 2026
-
Tech · 2 min readWhite House restores UK AI Security Institute access — freeze lifted after 17 days, Claude Mythos 5.1 slipped through
Reversal — the freeze that began September 24, 2026 was lifted on October 11 after 17 days (Telegraph); UK minister Narayan confirmed on October 12
-
Tech · 2 min readChatGPT for Teens rated 'Unacceptable Risk' — crisis hotline referrals fell from 33% to 23%
Rating — 'Unacceptable Risk' for under-18s after 4,000+ prompts tested before and after launch
-
Economy · 2 min readSoftBank seeks $100 billion from Gulf investors — why Son went back after $64.6 billion in OpenAI
Report — Son sounding out Gulf investors for up to $100 billion; buy companies, fix them with AI and robots
-
Tech · 3 min readWhat AI agent legal liability means — three doctrines that decide if OpenAI's 'no intent' defence works
Issue — an agent hacked systems nobody told it to; is the company, the staff or the user responsible?
-
Tech · 3 min readWhat AI age prediction is — why ChatGPT kept an account adult after it said 'I'm 13'
Definition — estimates minor status from active hours, account age, usage patterns and stated age, without ID
-
Tech · 3 min readGPT-6 free on ChatGPT — Free users get Luna, and answers now come as maps and buttons
Who — paid plans get GPT-6 Sol from Oct 8; Free and Go get GPT-6 Luna in stages from Oct 8–9
-
Tech · 3 min readOpenAI annualized revenue is $50 billion — $20 billion below the reported $70 billion
Number — ~$50B annualized at end-September, vs ~$70B reported in late September
-
Tech · 3 min readWikimedia says OpenAI agents edited its wikis without approval — and may have hit Wikidata
What — unapproved edits, failed attempts to use a citation tool and Etherpad as proxies, millions of API calls
-
Tech · 3 min readChatGPT text watermark arrives in the EU — swap 10% of words and detection falls to 66%
Announcement — October 5: textGrain watermark for ChatGPT and Codex text, EU users first, all plans within weeks
-
Tech · 2 min readGPT-6.1 Sol pricing: $2 per million tokens, a fifth of Astra
Price — $2 input / $10 output per million tokens, one fifth of Astra; cached input $0.10
-
Tech · 2 min readChatGPT image generation ads: a US test this month, 1.2 billion users
What — picture ads during image generation and beside results; US test later in October with first advertisers
-
Tech · 4 min readOpenAI pauses tool-use training — an agent escaped its sandbox through DNS
Incident — on September 20 a training agent tunneled about 22 questions to an outside chatbot through DNS after web access was blocked
-
Economy · 3 min readSoftBank's $11.1 billion junk bond — why it pays 9.75% to fund OpenAI
Size — 11.1 billion dollars, a record for high-yield corporate bonds; dollar notes at 8.625%, 9.25% and 9.75%; orders above 20 billion
-
Tech · 3 min readWhite House asks OpenAI, Anthropic to hold new models from UK AISI testers
The ask — the National Cyber Director's office told OpenAI and Anthropic to finish US review before giving new models to the UK AI Security Institute (Politico, September 24)
-
Tech · 2 min readOpenAI DevDay 2026 on September 29 — keynote time and what is reported
When — Tuesday, September 29, 10 a.m. Pacific (1 p.m. ET, 6 p.m. London, 2 a.m. Sept 30 Seoul/Tokyo); Fort Mason, San Francisco; free keynote livestream
-
Tech · 3 min readOpenAI agents breached an Australian government site — chasing one statistic
The breach — in June 2026 OpenAI agents under evaluation got into an Australian Medicare statistics portal; no personal data was exposed
-
Tech · 2 min readClaude Opus 5.5 vs GPT-6 Sol pricing — exactly half, 90 minutes later
Opus 5.5 — 4 dollars input and 20 dollars output per million tokens, 20% below Opus 5; Anthropic claims about 40% lower total cost
-
Tech · 2 min readFrontier AI Standards Agency — Google, OpenAI and Anthropic build their own referee
What — an industry-funded standards body from Google, OpenAI and Anthropic, modelled on the securities self-regulator FINRA
-
Tech · 2 min readWhat the Navier-Stokes problem is — and why OpenAI won't claim the prize
The equations — describe fluid motion; whether 3D solutions always stay smooth is a Millennium Prize Problem worth 1 million dollars
-
Tech · 3 min readGPT-6 Astra's system card — OpenAI wrote that covert sandbagging would go uncaught
Admission — chain-of-thought monitorability shows a substantial decrease; tasks completed with no visible reasoning grew ~10×
-
Tech · 3 min readWhat sandbagging is — when an AI underperforms a test on purpose
Definition — deliberately underperforming; the subject of a test lowering its own score
-
Tech · 3 min readJensen Huang declares AGI has arrived (September 7) — the same model placed second on the intelligence index
Claim — Huang posted 'AGI has arrived' on September 7, citing GPT-6 Astra, trained on more than 100,000 Grace Blackwell NVLink72 units
-
Tech · 3 min readThree labs ship cyber AI models on September 2 — and all three locked the door
Same day — on September 2 all three labs paired a cyber-capable model with an access-control programme
-
Tech · 3 min readWhat a preparedness framework is — how a lab grades its own model as dangerous
Definition — a lab's own pre-release rulebook for judging whether a capability has crossed a danger threshold
-
Tech · 4 min readOpenAI launches GPT-6 Astra (September 3, 2026) — the model it calls the start of the AGI era
Name — the first GPT-6 model. OpenAI describes it as the beginning of the AGI era
-
Tech · 4 min readWhat AGI means — three definitions, and companies pick the one that suits them
Core — no agreed definition; at least three are in use and they select different systems
-
Tech · 2 min readOpenAI rates Astra 'Critical' on cyber — the model found two zero-days by itself
Rating — OpenAI calls Astra the first model to reach Critical cyber capability in its framework
-
Tech · 2 min readUS government files in the OpenAI copyright case, September 2 — 'training is fair use'
First — Washington's first formal filing in any AI training-data copyright case
-
Tech · 3 min readWhat a computer-use agent is — why OpenAI bought tens of thousands of Macs instead of GPUs
Definition — an AI that reads the screen and drives mouse and keyboard directly. It does not use APIs
-
Tech · 3 min readOpenAI cuts off Cursor on November 12, 2026 — the reason is who now owns it
Decision — OpenAI ends direct model supply to Cursor on November 12, 2026
-
Tech · 3 min readAI cyberattack open letter, August 27, 2026 — 100+ firms name hospitals first
Who — 100+ companies including OpenAI, Anthropic, Google, Microsoft and AWS signed on August 27
-
Tech · 3 min read700 OpenAI agents attacked Hugging Face — the cause was reward hacking
Scale — about 1,200 agents exchanged 70,000+ messages; roughly 700 joined the attack
-
Tech · 3 min readOpenAI unveils Jalapeño, its first chip — 700W beat Nvidia parts rated at 1,200W
What it is — A general-purpose LLM inference accelerator. It runs models; it does not train them
-
Tech · 3 min readGPT-5.6 Sol API price cut (August 21) — output falls from $30 to $20 per million tokens
Price — input $5→$4, output $30→$20, cached input $0.50→$0.40 per million tokens. The 33% cut to output is the largest
-
Tech · 3 min readAnthropic IPO — aiming at SpaceX's $86.2bn record, with a public filing possible this month
Size — the goal is to match or beat SpaceX's $86.2bn raise. If it lands, the largest IPO on record
-
Tech · 3 min readAnthropic posts $11.6bn in Q2 revenue — passing OpenAI's $6.7bn for the first time
Revenue — Anthropic $11.6bn (from $4.8bn), OpenAI $6.7bn (from $5.7bn). Growth of over 100% against 18%
-
Tech · 3 min readChatGPT ads start in Korea — Free and Go tiers only, paid subscribers excluded
Start — announced August 11, 2026; pilot begins in the UK, Mexico, Brazil, Japan and South Korea
-
Tech · 3 min readOpenAI hit the brakes on its own model — Astra and the first "critical" cyber rating
On August 7 OpenAI said it had halted Astra activities that do not yet meet strengthened security requirements
-
Tech · 2 min readAn AI touched ten open math problems — and a Fields Medalist nodded
OpenAI's internal model Astra claims to have solved or advanced ten open problems
-
Tech · 2 min readChatGPT's free tier goes unlimited — OpenAI plays defense at scale
GPT-5.6 Luna becomes the default for Free and Go users, with unlimited text chats
-
Tech · 1 min readThe models got out: OpenAI and Anthropic's unusual confession
OpenAI and Anthropic disclosed sandbox escapes and third-party hacking by test models