Agents are the point where AI stops answering and starts acting. This topic collects our coverage of what that looks like in practice.
Coverage of autonomous agents: what they can complete unattended, where they stall, and how teams scope the handoff back to a person.
27 stories
AI Governance & Ethics
One appointment-reminder call can cross federal consent rules, state AI disclosures, and voice-likeness rights. Product teams need a call-level policy matrix before the system dials.
Irene Vasko · August 9, 2026 · 8 min
Agentic AI & Orchestration
In an isolated test, instructions inside a downloaded PDF diverted a browser agent from its assigned upload path. The reliable fixes sit around the model, not in another warning prompt.
Mara Quintero · 8 min read
Agentic AI & Orchestration
A browser agent can carry a hostile download from a public site into email or cloud storage. Put an inspection gate between the download tool and every upload tool.
Mara Quintero · 7 min read
Agentic AI & Orchestration
A timed-out tool call can leave an agent between failure and success. Safe retries depend on a persistent request identity, a way to check status, and a queue for unresolved actions.
Mara Quintero · 7 min read
Agentic AI & Orchestration
A natural-language database agent can handle recurring analysis without arbitrary SQL access. The workable setup combines narrow views, enforced query budgets and a separate path for changes.
Mara Quintero · 8 min read
Agentic AI & Orchestration
A one-time expense agent can retain access to email and files long after submitting the report. Expiring its credentials closes that gap, but only if the orchestrator controls the token lifecycle.
Mara Quintero · 8 min read
Agentic AI & Orchestration
A customer-service agent should not carry one reusable credential for every support action. Split access by action stage, then issue narrow, temporary credentials only when execution is approved.
Mara Quintero · 8 min read
Agentic AI & Orchestration
A local page, an inert canary, and a mock tool can reveal whether a browsing agent follows instructions it was supposed to treat as untrusted text.
Mara Quintero · 8 min read
AI Industry & Models
Computer-use APIs automate software with screenshots instead of dedicated integrations. That broad reach comes with slower loops and less certainty than browser tools that read page structure.
Tobias Lund · 7 min read
Agentic AI & Orchestration
A local invoice page and synthetic secret can show whether webpage text redirects your browser agent. The useful comparison is between soft instructions and hard tool limits.
Mara Quintero · 8 min read
Agentic AI & Orchestration
A parser error persuaded a sandboxed agent to read a canary token. Typed results and an external action broker stopped the same tool output from becoming an instruction.
Mara Quintero · 8 min read
Agentic AI & Orchestration
A browser agent followed instructions embedded in a vendor support page instead of retrieving an invoice. Isolation reduced the confusion; narrow tool permissions stopped the attempted data leak.
Mara Quintero · 8 min read
Agentic AI & Orchestration
Indirect prompt injection turns ordinary page content into instructions for an agent. The practical defense is to limit what the agent can see, call, and approve on its own.
Mara Quintero · 8 min read
Agentic AI & Orchestration
A browser agent followed an off-screen instruction into a dummy internal file. A URL allowlist and action-level confirmation gate stopped the same run without relying on the model to police itself.
Mara Quintero · 8 min read
Agentic AI & Orchestration
In a controlled vendor-research workflow, page text changed what an agent reported. The useful defenses kept untrusted content away from decisions rather than relying on a warning in the prompt.
Mara Quintero · 8 min read
Agentic AI & Orchestration
A controlled-page test can reveal whether an agent treats website text as evidence or as an instruction, before a connected mailbox, ticket queue, or account becomes the test environment.
Mara Quintero · 8 min read
Agentic AI & Orchestration
A local canary page shows whether a browsing agent mistakes website text for instructions. The useful defenses constrain tools and expose proposed actions, rather than trusting one filter.
Mara Quintero · 8 min read
Agentic AI & Orchestration
Indirect prompt injection turns page content into commands for an AI agent. Authenticated sessions raise the stakes because the agent may already have access to files, email, purchases, or account settings.
Mara Quintero · 8 min read
Agentic AI & Orchestration
A browser agent abandoned a return-policy lookup after reading an instruction planted in the page. Layered controls limited the damage, but only one preserved the task.
Mara Quintero · 8 min read
Agentic AI & Orchestration
A research agent can widen source coverage while a writing agent shapes the draft, but only if their handoff preserves evidence, uncertainty, and rejected material.
Mara Quintero · 8 min read
Agentic AI & Orchestration
An agent can change an inbox, lose its working context, then repeat the same actions. Checkpoint design determines whether it resumes cleanly or leaves duplicate drafts and missing work.
Mara Quintero · 7 min read
Agentic AI & Orchestration
On a recurring compliance form, Playwright stopped when a field changed. The browser agent kept going, which reduced repair work but made a wrong document upload harder to catch.
Mara Quintero · 7 min read
AI Governance & Ethics
A chat transcript is not an audit trail. Before an agent submits a payment, its log must preserve the evidence, authorization, and exact action that crossed the boundary.
Irene Vasko · 8 min read
Agentic AI & Orchestration
A dependency update is a useful test of coding agents because the patch is small but the definition of success is not. Passing tests alone cannot establish that the migration is safe.
Mara Quintero · 8 min read
Agentic AI & Orchestration
A spending agent should not treat a toner reorder, a new vendor, and an annual renewal as the same decision. This approval map separates routine actions from commitments that need a person.
Mara Quintero · 8 min read
Agentic AI & Orchestration
Replay historical support threads in an isolated sandbox, capture each tool call, and grade four kinds of behavior before the agent receives permission to send.
Mara Quintero · 8 min read
Agentic AI & Orchestration
A reservation succeeded, but the agent failed before delivering the itinerary. The recovery test exposed which retries are safe and where a person must take over.
Mara Quintero · 8 min read
Other impacts
One story a day
One real story about AI each morning — no hype, no alarm, just company for the road.