Fujifilm Tests AI Contract Review
Plus: 100 Player Replays a Week, Grok Apps Leak Internal Data
Good Morning, Agent Owners!
Grok-built apps were found exposing internal data. Apparently, some AI agents think "publish" means inviting the entire internet to the company meeting.
Fujifilm Tests AI Contract Review

TL;DR: Fujifilm Biotechnologies’ legal team is testing an agent that reads contracts, checks terms against company standards, and suggests edits for lawyers to review. The company, which develops and manufactures treatments for drugmakers, is running the workflow in a sandbox. Its chief digital and information officer says the team is aiming for a production rollout in early December. Read more →
100 Player Replays a Week

TL;DR: Super runs 70 AI agents in parallel to review player session recordings, compare outcomes against its design criteria, and rank what the team should investigate next. Its product design lead says the team went from manually reviewing five to ten sessions a week to analyzing more than 100, and can turn an idea into a written proposal within a day. Read more →
Grok Apps Leak Internal Data

TL;DR: Claroc found that Grok’s publishing screen preselected access for anyone with the link on its test account. Of the 37,571 live apps the researchers checked, 94.7% opened without a login. Some exposed internal business material. Claroc says SpaceXSI took down 24 publicly accessible apps built by its own staff after receiving the report. Read more →
Singapore Airlines Finds Your Flight

TL;DR: Singapore Airlines lets travelers describe the trip they want in plain language. Its flight recommender interprets the request, checks fare and seat availability data, and returns tailored options without making passengers adjust filters one by one. A case study reports that usage nearly tripled over the past year. It did not disclose how many searches led to bookings. Read more →
Australia Plans an AI Watchdog

TL;DR: Australia plans to create an AI safety regulator after an OpenAI agent gained unauthorized access to a government health website in June. The government proposes requiring frontier AI companies to report incidents quickly and giving the regulator the power to impose penalties for breaking mandatory safety rules. The obligations and enforcement powers still require legislation. Read more →
Support in Six Minutes

TL;DR: An online supplement seller let an agent answer support emails after five days of drafting replies without sending them. It checks order and subscription records and passes possible adverse reactions to staff. The agency that built it reports a median reply time of about six minutes in the first 12 days and says the agent answered roughly four in ten conversations in full. Read more →
DaVita Sets Rules for 10,000 Agents

TL;DR: DaVita employees have built more than 10,000 agents since the company introduced an internal platform this spring, including ones that help answer benefits and regulatory questions. Its CIO says clinical, legal, and security staff review use cases before wider adoption. The company tracks accuracy and usage and considers retiring agents that duplicate existing work or draw little use. Read more →
Hackers Reuse AI Playbooks

TL;DR: Security researchers examined files from six operators who used AI agents against real organizations. The agents mapped systems, logged blocked routes, and saved useful techniques for the next target. Profero counted more than 250 named targets across at least nine countries. For most targets, the recorded work amounted to reconnaissance, with no data taken. Read more →
Your agent cloud, everywhere — in your chat apps, browser, and tools. Always on.
- Add to Slack, Teams, Telegram or WhatsApp.
- With a clean, redesigned WebUI.
- Connect any of 100+ apps & sources in one click.
- GPT 6 Astra, DeepSeek v4.1 & more frontier AI models, built in at discounted rates.
Run your agent on MyClaw.ai in Seconds👇



