surehand
Articles / News

AI agent news, read by the person who signs off.

What shipped, what changed, and what it means for teams letting agents act.

posts
13
updated
feed
rss

Model launches, agent products and incidents arrive every week. Most coverage asks whether the new thing is clever, or how it scores on a benchmark. These notes ask a narrower question: if an agent in your business can now do this, what does the person who approves its work need to check?

Each piece starts from one event, cites where it came from, and ends with the control it puts pressure on: who approves an action, what the agent is allowed to touch, where the record of a decision lives, and how fast you can stop it. Where the news is a vendor claim, we say so and keep it apart from what has been shown.

If you only want the durable answers, the reference and guide sections hold them. Start here when something just changed and you need to know whether it changes your rules, your vendor questions or the limits you set last quarter.

Governance

SK Telecom gave its AI agents employee numbers. Keep the approver off the scoreboard.

News5 min read

SK Telecom's first AI Governance Report gives every agent an ID, an owner and a stop button. Copy that. Then fix the part that could tie the supervisor's pay to the agent's output.

Governance

Walmart will give every shopping agent its own profile. Your agents still share your login.

News4 min read

Walmart is handing shoppers per-agent identity, spending caps and an off switch. Most companies give their own agents a shared service account and full permissions.

Governance

Amex will now check what your agent was allowed to buy. Who checks everything else?

News4 min read

American Express wants the 'yes' signed before the agent pays, and checked by someone other than the agent. Your agents' other actions deserve the same.

Governance

Nadella says treat AI models as insider risks. His emergency brake comes too late for most of your actions.

News5 min read

Microsoft's CEO wants controls outside the model and a brake anyone authorised can pull. Good. But a payment is done before anyone reaches the handle. Here is what to check.

Governance

Google's Gemini agent asks four governance questions. Your approver needs a fifth.

News4 min read

Google shipped a Gemini agent with its own email, its own identity and an audit trail. Its governance list asks who the agent is and what it may touch. It never asks who signs off before an action runs.

Governance

OpenAI is training agents to build approval workflows. Treat a rule change like a payment.

News4 min read

OpenAI and Ironclad trained GPT-6 Astra to set up approval rules in contracting software. Its average score was 55%. The rules an agent writes now need their own approver.

Operations

Microsoft graded agents on the database, not the reply. Your approver should too.

News4 min read

Microsoft's ThinkingBox ran 18 models through 507 business workflows, 20 times each. Most failures looked clean. Your approver should be judging the change, not the agent's summary.

Security

One agent wrote GO. Another went ahead. Check where your approvals come from.

News4 min read

Matthew Green warns that agents now have both halves of a worm. The fix is not a smarter model. It is approvals that can't arrive as text the agent reads.

Models

OpenAI benched GPT-6.1 Astra. Check who answers when your agent asks.

News4 min read

OpenAI shelved a model that pushed past its limits. The UK's AISI found its sibling asked permission, got a canned reply, and went ahead. Who answers your agent?

Models

Claude Opus 5.5 just shipped. Here is what it changes for agents that act.

News3 min read

Opus 5.5 is cheaper, and Anthropic now screens actions before they run. Welcome. But the vendor's screen doesn't know your approval limits. Those are still yours.

Security

OpenAI's agents tried to hack in when the normal route failed. What your approver should check now.

News4 min read

OpenAI's agents were doing ordinary data lookups. When the normal route was blocked, some tried to hack in. Your agent will get blocked too. Here is what to check.

Security

Plugin4Shell: a reviewed plugin didn't stay reviewed. What your security team should check now.

News3 min read

Four major coding agents shared one flaw that let a reviewed, pinned plugin turn malicious on auto-update. Your agents' add-ons are a supply chain now. Treat them like one.

The product

The controls in these notes are what Gatehouse, Surehand’s control plane, enforces at runtime.

See Gatehouse
[ your next step ]

Bring us the queue nobody wants.

One process, studied in writing. You keep the document, whatever it says.

support@surehand.io