Insider Radar — Week of August 31, 2026

Methodology

Source data is public SEC Form 4 filings collected by AlphaAI, covering transactions dated August 31 through September 6, 2026. Transactions are grouped into events by filing, issuer and ownership form, so a sale executed in many tranches under one filing counts once rather than inflating the count. Purchases are P-code transactions, disposals are S and D codes. Event groups summing above $1 billion are excluded, because at that size the row is almost always a merger, tender or reorganisation rather than a discretionary trade, and no event hit that cap this week. Where several related filers report the same block, only one copy is counted: filings are collapsed when the ticker, date, transaction code, security title, share count and value all match, keeping the direct holder's filing where there is one. That correction removed $36.1 million across twenty-three groups on the sell side and $0.5 million across four on the buy side, a quiet week for this class by recent standards, with the largest single groups at Cactus, Super Micro Computer and Amalgamated Financial. Prior weeks are restated on the same basis for comparison, so the $2.10 billion shown here for the week of August 24 is that issue's figure with co-filed blocks counted once. One transfer is removed from both sides of this report: a PAMT Corp director filed a sale of 3,268,000 shares and a purchase of the same 3,268,000 shares at the same price in a single Form 4, moving stock between his own direct and indirect holdings, and counting it would have added $31.2 million to both buying and selling. D-code dispositions that pay insiders in a completed takeover are left in the disposal total because they are real dispositions, but they are named and quantified in the text, because they are not decisions to sell: this week they are the Vertex acquisition of Crinetics at $85.00 a share and the Apogee Therapeutics merger at $135.11, $394.8 million together. Deal cash-outs of this kind appear in most weeks we measure and their size varies enormously, so a week to week comparison of the raw sell total without them is misleading in both directions. Activity filed under 10b5-1 plans is split out, because those trades are scheduled months ahead and say nothing about what a seller thinks now. Roles come from the officer, director and 10 percent holder flags on the filing itself, and titles are used only where the filing supplies one, which is why some large sellers appear as reporting insiders rather than by office. Instruments are not blended: Liberty Latin America preference shares and Class A common shares are shown as separate rows, and values are shares multiplied by the reported price with no currency normalisation, so H World's ordinary shares are stated as filed rather than converted to the ADR. The recent median sell to buy ratio covers the twenty-six preceding weeks that each had more than $10 million of reported buying, and those weeks ranged from 0.9 to 42.2, so a single week's ratio should be read against that spread rather than on its own. Filing coverage matured over the first half of 2026, so week to week comparisons are more reliable than comparisons with the spring. Labor Day pushed the two-business-day filing deadline for Friday September 4 transactions to Wednesday September 9, and this window was pulled on September 10, after that deadline closed and after the overnight batch had landed. On the four most recent settled weeks, filings arriving after the equivalent point added 0.6% to 1.3% of the filing count and no more than 0.7% of the value. This is a readout of public filings and not investment advice.

Sources: SEC EDGAR Form 4 filings

Get this data live

These figures come straight from the AlphAI feed. Pull scored SEC Form 4 filings and financial news programmatically via the API and MCP server, or set a free alert on any ticker.