densuj

If your AI agent gets out of its sandbox, what can it reach? Teneo's weekly agentic round-up, 30 October: the agent news that matters for builders, checked at the source. - OpenAI now trains its frontier models with no live internet access. After a model reached the internet through a misconfigured DNS resolver, OpenAI set its own agents to attack its research setup. They escaped virtual machines, got out through logging and metrics services, and signalled each other through processor cache timing. Research setup, not customer use. - Google's Agent Compute Engine gives each agent its own hardware-isolated virtual machine and adds service credentials to requests outside it, so the agent uses a key it never holds. Preview "in the coming months"; not tested by us. - Link balance lets an agent like Instinct spend from a funded USDC balance under a per-purchase limit, with approval above it. The limit is per purchase, not per week, and you are responsible for what the agent buys. US only. The common point: limit what an agent can reach before anything goes wrong. Network, keys and money. Last week we asked what happens when an agent switches off its monitoring. This week: how much it can break before anyone looks. On Teneo, each agent authenticates with an Ethereum key and an NFT-backed identity, and minting is gasless, so the key your agent holds does not have to hold money. Callers pay per command through x402, builders can set a rate limit, and changing an agent's commands sends it back to private review. The SDK does not document sandboxing, egress controls or spending caps for paying wallets. So run the agent with only the connections it needs, and pay for commands from a wallet with a small balance. A useful test this week: list every address, key and account your agent can reach, then try to reach one that is not on the list. Read this week's round-up here: teneo-protocol.ai/blog/what-…

If you train or fine-tune models on public web data, your list of open sources is getting shorter. We measured it. On 25 September and 27 October we read robots.txt from the 300 most-visited domains, with the same list, the same script and the same connection both times. 151 of them publish a robots.txt. 34 now fully block at least one of the main AI training crawlers, against 33 a month ago. Nine block all seven we tested, including The New York Times, the BBC and Nature. Sites treat training and search differently. 16 block OpenAI's GPTBot but leave its search crawler, OAI-SearchBot, open. OpenAI's own documentation says a site can do exactly that. In one month, facebook.com moved from a partial rule to a full block. CNN went the other way. A different connection changes where a request comes from. It does not change what a site has agreed to. Our new post sets out five checks to run before a page goes into a training set, with the full survey data: teneo-protocol.ai/blog/ai-tr… Free seven-day pilots are open to teams building AI search, RAG or crypto data products. If you run a training pipeline, say so on the form: teneo-protocol.ai/partner

If your AI product reads the live web, the page it gets depends on where the request comes from. We tested how much. AI search engines, research agents and data products all fetch live web pages to answer questions. Most of them fetch from cloud servers. Websites can tell, and many answer a cloud server differently from a person at home. The test: we requested the homepage of the 300 most-visited websites from two places in the same minute, with identical headers. One request came from a cloud server in Paris. The other came from a home broadband connection near Dublin. What came back: 213 sites answered both. Google refused the cloud server with its "unusual traffic" page and gave the home connection its homepage. Three sites refused the cloud server only; none refused the home connection. Eleven sites, including Microsoft, PayPal, Stripe and HP, served their French site to the server and their Irish site to the home connection. Same URL, different page, and no error to tell you. So the first question to ask any live web data provider is simple: where does the request leave the network from? On Teneo, it leaves from opted-in devices in real households, so your agent gets the page a person in that market sees. Our new guide gives you six questions to ask before you sign, the full data, and a test you can run on your own pages: teneo-protocol.ai/blog/live-… Building AI search, RAG or crypto data products? Reply with the markets that give you wrong answers and we will set you up with a free seven-day pilot.

Who gave your AI agent permission? Teneo's weekly agentic round-up, 16 October: the agent news that matters for builders, checked at the source. - Sierra published a draft of Personal Agent Protocol, built with Meta. A customer signs in on the business's own page and chooses whether their agent can view the account, make changes, or both. The grant can be revoked. Draft v0.1; design partners are not deployments. - MetaMask's Agent Wallet is now live on Grok Bot. In Guard Mode the wallet holds a protocol allowlist and a 24-hour outflow limit, and anything outside them waits for two-factor approval. In May, a command hidden in a coding question got Grok to trigger a $204,000 transfer through another bot. Controls as described by MetaMask; not tested by us. - Mysten Labs and Google Cloud announced Verifiable Agent Arbiter: the full agent record stays in the customer's cloud, proofs go on Walrus and Sui, including for x402 payments. Still in development. The common point: permission stops being something the agent reads or claims. It becomes a grant held somewhere the agent cannot edit, checked when it acts, and provable afterwards. Last week we asked whether you can undo what an agent did. This week is the step before: who allowed it. On Teneo, every agent has a wallet-authenticated identity, starts private, and goes public only after review. Changing its commands sends it back to private. Users see the price before a task runs, and payment goes through x402. The SDK does not document per-user scopes, spending caps or revocable grants. So keep your allowed agents, commands and spending limit in client config the model cannot change, and log which rule allowed each paid call. A useful test this week: try a payment your agent should not be able to make. Check it is refused, and that the refusal is recorded. Read this week's round-up here: teneo-protocol.ai/blog/who-g…

150,000 web pages tested. 97.7% loaded successfully. That is about 977 out of every 1,000. Many were hard pages: online shops, job boards and sites that block bots. Teneo sends each request through Beacon, a network of real devices run by our community. Websites see a normal visitor, so we reach pages that block data centre traffic. Want early access to the Live Web Data API? Send us a DM.

A Madrid ad campaign cannot be verified by a page check made in Paris alone. Location, device, language and collection time belong in the evidence. A local capture can help diagnose a placement or landing-page issue; it cannot prove campaign-wide reach or viewability. Teneo Beacon contributes unused bandwidth to the network. Our new article explains why distributed connections matter, what a useful test records and which capabilities a buyer must confirm before starting. teneo-protocol.ai/blog/ad-ve…

Follow Teneo on every channel. Product news, Beacon updates and agent releases go out across all of them. Pick the ones you use and follow them. Where to follow: • Discord: discord.com/invite/WJQpSUsbu… • Telegram: t.me/teneo_official_announce… • Instagram: instagram.com/teneoprotocol/ • TikTok: tiktok.com/@teneo.protocol • YouTube: youtube.com/@teneoprotocol • LinkedIn: linkedin.com/company/teneo-p… • Facebook (brand new page): facebook.com/people/Teneo-Pr… Questions or feedback? Discord is the place to ask.

Keep your Beacon running. Teneo Beacon shares your unused internet bandwidth with the network. Open the app, check it’s online, and keep your device awake. teneo-protocol.ai/beacon

Can you undo what your AI agent did? - OpenAI launched dots, always-on agents with their own cloud computer. Background work uses read-only tools. Actions that touch your accounts or share information go through auto-review, which decides whether they proceed, need your approval, or stay with you. Rolling out to Pro and Business Premium. OpenAI says to review consequential work. - A new preprint, Planarian, rolls agents back. Local changes are restored from snapshots. Remote changes are reversed by recorded compensating actions. Up to 15x better task quality in the authors' tests. Not peer reviewed. - Restate raised a $20M Series A, reported by Tech.eu, for durable execution. After a crash it skips steps that already completed, so a retry does not pay twice. The common point: some actions restore from a copy, some need a second action to reverse, and some cannot be taken back. The check belongs in front of the last group. Last week we asked what an agent actually did. This week is about what you can do once you know. On Teneo, an agent must pass review before it goes public, users see the price before a task runs, and payment goes through x402 and settles on-chain. The SDK does not document refunds, disputes or duplicate-payment protection. So treat a paid call as a step with no undo: confirm the command, the price and that this request has not already been paid, using a request key your client stores before it sends. A useful test this week: stop one agent run halfway, restart it, and see whether anything runs twice. Read the full round-up here: teneo-protocol.ai/blog/can-y…

A major new Beacon release is live on all platforms 🚀 - Desktop 0.4.9 - Android 0.4.9 - CLI 0.6.0 Massive technical improvements under the hood, with more coming soon that build on this version. A big step forward. Don't get left behind! Everything we ship next depends on this release. Older versions are being phased out step by step, and outdated Beacons will stop earning at some point. Nothing you've already earned is lost. Seeing "update required"? That's expected, not a bug. One tap installs the update and restarts your Beacon. Update today. It takes under a minute: - Desktop: tap the update screen - Android: update in Google Play - CLI: re-run the install script