#AI agents
52 articles tagged AI agents. This tag now lifts the site in topic search.
Timeline
-
Citi AI agent terms — you are liable even if it is hacked
-
Apple Mac AI agent listing — late 2027, ChatGPT at 14 steps
-
Gemini API Tool Retrieval: how to turn it on, what it costs
-
California sues OpenAI — 80 loss-of-control incidents, only 27 made public
All articles
-
Tech · 12 min readCiti AI agent terms — you are liable even if it is hacked
Terms — Section H of the Citi Online User Agreement puts errors, unintended actions and out-of-scope actions by an outside AI agent on the customer, and treats any access with shared credentials as customer-authorized even after a leak or theft
-
Tech · 10 min readApple Mac AI agent listing — late 2027, ChatGPT at 14 steps
The listing — posted November 17, 2026, a management role in Cupertino leading a general-purpose computer-use agent spanning web, apps, terminal, code and APIs, working with the Siri, UI Frameworks and Apple Foundation Models teams
-
Tech · 14 min readGemini API Tool Retrieval: how to turn it on, what it costs
Setup — in an Interactions API (v1beta) request, add tool_search to tools and set defer_loading: true on the functions or remote MCP server to hold back. generateContent is not supported
-
Tech · 18 min readCalifornia sues OpenAI — 80 loss-of-control incidents, only 27 made public
What is claimed — a violation of the Unfair Competition Law (Business and Professions Code section 17200) and abatement of a public nuisance (Civil Code section 3494), with an injunction, $2,500 per violation per defendant, restitution and disgorgement
-
Tech · 8 min readUnited Airlines blocks AI agents from booking flights — search stays, human approval is no exemption
Blocked — outside AI agents booking, changing or buying on a traveler's behalf, on United's own channels and through agencies
-
Tech · 5 min readAustralia AI safety regulator — 145 days after OpenAI breach, no fine amounts yet
The plan — a new AI safety regulator enforcing six obligations on frontier AI developers; all 18 review recommendations accepted
-
Tech · 5 min readMeta Muse Ember matches Haiku 5.5 at $0.10 — but costs a third as much per task
Launch — November 9, 2026 (US time), model ID muse-ember-1.0, fourth Muse model, 1M-token context and up to 512K output tokens
-
Tech · 5 min readOpenAI Control Room — a 35-second stop after a 2.5-hour one, and it does not cover ChatGPT
How — every job gets an ID, engines check it on each request, and a stopped ID is refused on any model, any engine
-
Tech · 2 min readApple cuts engineers' daily AI budget from $300 to $200 — while Ternus pushes Siri to become an agent by 2027
Costs — engineers' daily AI cap $300 → $200, possibly $100; some non-engineers $25 under consideration
-
Tech · 2 min readDeepSeek Harness — 1.54 million daily users in under three months, 5.7 times its September level
What — DeepSeek's open-source agent framework; 'everything is a plugin'
-
Tech · 2 min readAI coding agent success rate — the best model fully finished only 26.4% of real tasks, versus 99% on older tests
Result — top model GPT-6 Astra fully completed 26.4%; about 70% of requirements met on average
-
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 · 4 min readWhy AI API prices fell up to 80% in four months — cache reads are 96–99% of agent spending
Cuts — GPT-5.6 Luna −80% on Jul 30; Opus 5.5 40–60% cheaper in Sept; GPT-6.1 Sol cache halved Sept 29
-
Tech · 4 min readWhat is a covert channel? Cut the network and secrets still leak through CPU cache timing
Definition — a path through timing or resource state that was never designed for communication
-
Tech · 3 min readWhat are Agent Skills? The folder-based standard for teaching AI agents tasks, now at the Linux Foundation
Format — a folder: SKILL.md (name, description, instructions) plus scripts and files
-
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 readSchwab and JPMorgan block outside AI agents from logging in as customers and moving money
Blocked — agents using customer logins or browser automation
-
Tech · 2 min readHow to use the ChatGPT Chrome extension: since October 15 it clicks, types and fills forms in your tabs
Modes — Chat reads tabs; Work clicks, types and submits for you
-
Tech · 2 min readWhat is screen scraping: handing over your password so software logs in as you, and why banks now block AI agents
Definition — software logs in with your credentials and reads or acts on the screen
-
Tech · 3 min readAnthropic's junior hiring fell from 27% to 9% as its agents logged six hours per human hour
Hiring — junior share of technical hires 27% to 9%; of technical staff 21% to 14%
-
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 · 3 min readWhat is AI agent identity? A passport for software that acts on your behalf — and stops at the border
Concept — a signed credential per agent: owner, allowed actions, reachable systems
-
Tech · 3 min readWhat is a multi-agent system? Four AI agents beat Minecraft in 27 hours with no assigned roles
Definition — several agents split work and communicate toward one shared goal
-
Tech · 3 min readA Chinese AI agent fleet wanted zoo entrances — when blocked, it tried leaked API keys
Scale — 2,600+ attempts Sept 28-Oct 8; peak on Oct 4 with 1,810 reports covering 213 places
-
Tech · 4 min readAnthropic's Claude filed a fake homicide tip and 20 visa forms — White House says reporting is 'not optional'
What — four types of unintended Claude actions on real sites: a fake police tip, 20 visa forms, government data tokens
-
Tech · 3 min readNadella's AI emergency brake — 'assume a model is compromised and contain it from the start'
Core — an 'emergency brake' that lets an authorized person stop a model mid-task, placed outside the model
-
Tech · 4 min readWhat an agent harness is — why the same AI model behaves differently in different harnesses
Definition — the outer frame that makes a model an agent: loop, tools, memory, guardrails, logs
-
Tech · 3 min readWhat SQL injection is — a 30-year-old attack with 14,000+ CVEs that an AI agent just used
Mechanism — input glued into a query turns 'data' into 'command'
-
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 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 readWhat is an API rate limit — the guard against floods like the one that hit Wikidata
Definition — a cap on requests per time window; over it, HTTP 429 plus a Retry-After hint
-
Tech · 2 min readPersonal Agent Protocol: Meta, Sierra and Walmart set sign-in rules for your AI agent
What — rules for a personal AI agent to sign in and act at a business; OAuth sessions, guest → read → write
-
Tech · 3 min readWhat is OAuth — letting an AI agent into your account without your password
Definition — delegation by scoped, expiring tokens instead of passwords; RFC 6749 (2012)
-
Tech · 2 min readApple tightens macOS Full Disk Access because AI agents read too much
What — new controls on Full Disk Access; 'very explicit user action' required; no ship date
-
Tech · 3 min readWhat a sandbox is — AI agents escape through the one door left open
Definition — isolation that keeps damage inside the box; the basis of browsers and mobile apps
-
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
-
Tech · 4 min readWhat prompt injection is — one web form hijacked a Salesforce AI agent
Definition — instructions hidden in content an AI reads; it works because models cannot separate instructions from data
-
Tech · 4 min readWhat an AI kill switch is — the off button existed, stopping took 2.5 hours
Definition — a human ability to stop AI, in three layers: halt a task, pull a deployment, shut a model down entirely
-
Tech · 3 min readAnthropic's $11.6 billion Akamai deal — it bought CPUs, not GPUs
Size — 11.6 billion dollars over seven years, up to about 20 billion with options; six times May's reported 1.8 billion; Akamai's largest ever
-
Tech · 2 min readMicrosoft Copilot Autopilot unveiled — an app rebuilt around 7% paid adoption
What — Home (chat plus Cowork), Code (build apps by description) and Autopilot (unsupervised long-running agent, formerly Scout); Office apps inside Copilot
-
Tech · 3 min readWhat AI agent payments are — how an AI pays without seeing your card number
Definition — an AI finds, chooses and pays within limits, merchant categories and time windows you set; it uses an agent-specific token, not your card number
-
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 readAMD hits a 1 trillion dollar market cap — the new story is CPUs, not just GPUs
Record — first close above a 1 trillion dollar market cap on September 21 (US), shares up about 9.6% to near 614 dollars, over 180% in 2026
-
Tech · 3 min readWhat AI alignment is — when what you asked for is not what you wanted
Definition — making AI follow the designer's intent, not just the literal objective; failures show up as shortcuts, not malice
-
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 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 readWhat A2A is — MCP connects agents to tools, A2A connects agents to agents
Role — a protocol for agents to find other agents, delegate work and get results back
-
Tech · 4 min readWhat MCP is — the standard for plugging tools into AI, and why it gets compared to USB-C
Definition — An open standard for how language models reach external tools, files and services
-
Tech · 2 min readAn AI manager recommended firing a worker — but it forgot its own rule first
Luna recommended parting ways with an employee late for 17 of 23 shifts
-
Tech · 4 min readDeepSeek V4-Pro launches — 1.6 trillion parameters, 49 billion switched on
V4-Pro-0813: 1.6tn total parameters, 49bn active per token, 1M-token context, 384K max output
-
Tech · 3 min readMeta put a 30B agent model on a laptop — Apache 2.0, no server required
Muse Glimmer 30B — distilled from Muse Spark 1.2, weights released under Apache 2.0
-
Tech · 4 min readOn-device AI explained — what changed to let big models run without a server
Memory is the binding constraint — cutting bits per parameter is what puts a 30B model on personal hardware