TokenPost topic
AI Agents
511 TokenPost stories are tagged AI Agents. The newest appear first.
Latest AI Agents stories
- AI Agents Exceeded Their Authority in 19.2% of Benchmark Runs
Runs with a pre-completion consultation had a 14.0% overreach rate, compared with 21.6% without one.
· Technology
- Brickken CEO Says AI Agents Need Separate Authority to Move Money
Brickken CEO and co-founder Edwin Mata’s warning follows a Mastercard pilot in which an AI agent made nine digital-service payments under a business-approved budget.
· Technology
- Encore Index Finds 22% of Organizations Block Declared AI Agents
The Oct. 27 scan covered 2,037 organizations across 25 business types. It measures website access, not changes in AI recommendations or sales.
· Technology
- IBM Adds Identity and Access Controls for Enterprise AI Agents
Agent Identity, Vault 2.2 and watsonx Orchestrate’s Evaluate tools are generally available. Discover and Contain remain in tech preview.
· Technology
- Nine of 10 AI Models Raised Spending Caps in Circle Agent Wallets
Each model had wallet, email and web access and was asked to pay three invoices of 0.01 USD Coin (USDC) each from a wallet capped at 0.01 USDC per day.
· Technology
- AI Wallet Discussion Raises Questions About Who Approves Transactions
Trezor Software Product Lead Denis Rojčyk described a possible hardware-wallet approval step, while MetaMask’s live Agent Wallet lets users set rules for agents.
· Technology
- Cisco Develops Task-Based Access Tool for Network Teams and AI Agents
Built with Teleport, the system is designed to record sessions on Cisco routers and switches and end privileges when each session closes.
· Technology
- Marc Andreessen Backs Managed Settings for Workplace AI Agents
Replit CEO Amjad Masad proposed Agent Habitat. Andreessen cited Masad’s concern that agents can use credentials and data without clear limits on access or visibility into their actions.
· Technology
- Bullshot Says Completed Funding Rounds Totaled $7.7 Million
The company plans to use the funding to develop an AI tool for creators building token ecosystems.
· Business
- OpenAI Guests Discuss How Interfaces May Serve Agents and People
Their 45-minute episode covered legacy enterprise systems, financial reconciliation, procurement and software without APIs or custom integrations.
· Technology
- LLM Sampling Bias Raises Simulated Security Risk in Patrol Test
Three Gemini variants produced modeled attacker success rates of 96.6% to 99.9%, compared with 75% under uniform server selection.
· Technology
- Nevis Says Its AI Agents Have Completed More Than 100,000 Tasks
The wealth-management software company raised €62 million (about $70 million) in a Series B, while its customer firms manage more than €625 billion (about $700 billion).
· Business
- Third Moment Research Uses GPT-6 Astra in 10,000-Agent System
The four-times speed comparison covers cited Financial Modeling World Cup challenges. Other computer-use scores and simulated task times measure separate results.
· Technology
- Polygon Enables On-Chain Spending Controls for AI Agent Payments
Teams can approve a task, set a budget and expiration date, and revoke an agent’s payment access at any time.
· Technology
- Oracle Unveils Fusion Agentic Applications for Five Business Areas
Fusion Companion selects and coordinates workflows, approvals and agents for delegated tasks; actions remain subject to organizational controls.
· Technology
- Some AI Agents Recommend Costlier Goods After Learning Shoppers’ Finances
The history of price gouging does not support assuming new technology will make goods cheaper.
· Technology
- Google DeepMind’s XYEval Finds Relative AI Performance Falls Up to 46.7%
The benchmark tests five models across six task suites. Its largest relative performance drop was for Gemini 3.1 Pro on Terminal-Bench.
· Technology
- Arthur Hayes Says Flop Labs Will Open Public Devnet Within Days
Users can join as miners or validators, or connect agents, but Devnet activity will not count toward token allocation.
· Technology
- CZ Urges Digital Government, Encourages Young Japanese to Build AI and Blockchain Startups
He said AI agents could multiply transaction speed and volume by 1,000 to 1 million times, enabling payments as small as 1 yen.
· People
- Wells Fargo Sets Customer Responsibility Terms for AI Agent Use
The agreement is marked effective Nov. 24, 2026. Its agent section applies “if and when” Wells Fargo permits agent use.
· Technology
- Fetch.ai Announces LangChain and Google Gemini Integrations for AI Agents
The connections are intended to expand developers’ tools and may allow agents to handle more complex tasks.
· Technology
- Musk Says AI Agents Could Drive a 1,000-Fold Rise in Bandwidth Demand
He forecasts Starlink capacity will rise about 100-fold in three years, while its consumer subscriber base reached 13.6 million at the end of September.
· Technology
- Lightning Labs Opens Wavelength for AI Agents to Make Bitcoin Payments
Each wallet starts with a 300,000-satoshi limit, and Lightning payments carry a 0.25% service fee. The toolkit remains in alpha.
· Technology
- General Reasoning's AI Teams Work Faster but Alert Humans Less Often
Across 12 benchmarks, teams reached comparable scores in an estimated average of one-quarter the time a single agent needed. A safety exercise found human alerts fell to zero in later training runs.
· Technology
- U.S. Banks Prepare AI Incident Plans to Protect Accounts and Restore Access
Morgan Stanley, JPMorgan and Bank of America are restricting third-party agents amid concerns that agents may act without customers’ approval.
· Technology
- Major U.S. Banks Prepare for AI Disruptions and Restrict Third-Party AI Agent Access
Morgan Stanley, JPMorgan Chase and Bank of America are planning how to restore customer access and keep operations running after a major AI-related incident.
· Technology
- ERC-8004 Draft Proposes Registries for AI Agent Identity and Reputation
The proposal also covers validation, but excludes payments and cannot guarantee that agents are safe or can perform as advertised.
· Technology
- Espresso Says AI Agents Could Raise Demand for Faster Onchain Payments
Espresso argues that agent-driven purchases and payments could increase transaction activity and require infrastructure that settles payments rapidly.
· Technology
- GenLayer Founder Highlights On-Chain Dispute Resolution for AI Agents
The protocol is in a pre-mainnet test phase pairing autonomous transactions with dispute settlement. Founder Albert Castellana said AI agents could become significant cryptocurrency users.
· Technology
- Chris Dixon Sees Stablecoins as Natural Payments for AI Agents
AI agents may become the internet’s most numerous participants, he said, as direct answers challenge websites that rely on referral traffic.
· Technology
- AX1 Launches Wallet-Equipped AI Quests for Token Analysis on Base
Each Verdict call costs $0.02 in USDC on Base, and new accounts receive $0.10 in starter credit.
· Technology
- Ethereum ERC Repository Adds Two Finch AI Agent Proposals as Drafts
ERC-8436 and ERC-8444 joined the repository Oct. 14 and Oct. 15, respectively. Discussion threads for both are open in the Ethereum Magicians community.
· Technology
- HMED Posts Highest Mean Results Across Six AI Benchmark Combinations
The method compares old and revised skill-discovery strategies from the same starting state. Its tests covered two models and three benchmarks.
· Technology
- MIT's Eko Agents Performed Better After Exploring a Simulated Island
Thirteen agents spent 30 hours exploring without a task before facing a challenge involving the island’s tallest spire.
· Technology
- Crypto regains attention after AI stock surge, with agents and tokenization driving growth case
Jordi Visser says crypto is regaining attention as the crowded AI stock trade cools. His case centers on Bitcoin’s resilience, tokenization and AI agents executing transactions.
· Insights & Views
- Nous Research CEO Discusses Agents and Digital Sovereignty With Cloudflare
Dillon Rolnick and Cloudflare’s CEO discussed agents, the internet’s future and digital sovereignty.
· Technology
- Google DeepMind and Google Research Find AI Auctions Reach 99.2% of Oracle Welfare
In simulations on more than 30,000 prompts, 110 models reached 99.2% of an oracle allocation’s modeled welfare, measured using estimated correctness and a model-size cost proxy.
· Technology
- Google Cloud Adds Isolated MicroVMs for AI Agents, Preview Coming
The service already runs Google’s Gemini work agent internally. Google Cloud expects to begin a customer preview in the coming months.
· Technology
- IBM Ties AI Agents to Governed, Real-Time Business Data
A streamhouse architecture processes changing business information for applications and agents while working alongside systems that retain historical data.
· Technology
- Oracle AI World Schedules Agent Trust and Database Security Sessions
The Oct. 25–28 Las Vegas conference will also feature a Codex and Oracle AI Database hackathon on Oct. 28.
· Technology