Articles
Synthetic data is not one training policy Studies of recursive training reach different outcomes because “uses synthetic data” hides the decisive choice: whether generated examples replace the original corpus or accumulate beside it. Specify retention and sampling before judging the risk. Every agent retry needs the same operation ID A timeout does not tell an agent whether its tool call failed. Retrying without the original operation ID can repeat the effect, so idempotency belongs in the tool contract before an agent reaches production. Your backtest knows more than your system did A chronological train/test split can still leak the future. Historical evaluation must reconstruct the information available at each decision, including late arrivals and corrections, before its score can support a deployment decision. MCP moved client identity to a URL. The allow-list is yours to write. The July 2026 specification deprecates dynamic client registration in favour of a metadata document the client hosts itself. That makes the calling application identifiable for the first time, and it makes deciding which applications are acceptable a decision nobody in most organisations currently owns. The output cannot tell you whether the mechanism was ever there The Federal Trade Commission finalised orders on 27 August 2026 against three firms that sold an advertising product on a capability the agency says it never had. No customer could have found that by examining what they received, and that is the ordinary case in AI procurement rather than the strange one. Agent memory is a new system of record, and it arrived as a feature flag Persistent memory turns an assistant into something that writes its own records. Those entries are model-authored paraphrases with no schema, no owner and no expiry, and they carry the obligations of the material they summarise without carrying any of its controls. An AI-drafted control profile describes your documents, not your controls NIST has published worked prompts for drafting a Cybersecurity Framework profile from an organisation's own artefacts. A model given a folder of documents can only report what those documents assert, which leaves the note recording where the evidence ran thin as the part with assurance value. Only the questions where they disagree are deciding your model choice A comparison between two systems on the same evaluation set is carried entirely by the items the two answer differently, and that count is usually in the tens. Recent work reports that many published pairwise rankings are not resolved at conventional significance and power. Autonomy is not the risk in an agent. Irreversibility is. Review boards keep asking how autonomous an agent should be, and the question has no answer, because autonomy is not a quantity anyone measures. The answerable question is which actions leave effects nobody can take back — and that is a fact about the system around the model, not about the model. The architecture outlives the team that built it Even the best-resourced laboratories are losing senior people faster than they were three years ago. If retention is difficult there, an enterprise AI programme has to be designed on the assumption that the people who built it will not be there to explain it. The NIST documentation draft requires one field. Ask for the profile. NIST published the initial public draft of its AI dataset and model documentation templates in July 2026. Conformity to the base template requires exactly one populated field, and everything an enterprise buyer wants to read sits behind a profile — a document any interested party can write, including the buyer. Confirmation is not a mitigation. It is the classification. The MHRA's guidance of 29 July 2026 leaves ambient scribing tools outside medical device regulation only while their outputs restate what was said and a clinician confirms them. Both conditions are held in place by product decisions, and both can be undone by an ordinary feature release.
Bring us the question
Something here already on your desk?
That is the conversation we are best at. Thirty minutes, a written summary, no obligation.