(Read more: How to kill high-frequency trading)
Schneiderman has begun meetings with the exchanges and alternative trading venues to discuss reforms, according to a person familiar with the situation.
A spokeswoman for the New York Stock Exchange declined comment. A Nasdaq spokesman did not immediately return a call for comment.
Among the practices Schneiderman called into question were "co-location," which allows firms who pay a fee—typically thousands of dollars a month—to locate their computer servers within the exchanges' data centers.
(Read more: High frequency looking for IPO cash)
Co-location reduces by milliseconds the time it takes to transmit, long enough for "predatory" high-speed traders to benefit and for the markets to suffer.
For instance, he said, the traders look for arbitrage opportunities between and among venues to capture momentary differences in stock prices.
The firms also artificially inflate prices, he said, by detecting a big trade from an institutional investor and positioning themselves on the other side.
Institutional investors have been forced to develop strategies to hide their orders from these traders, such as by routing the orders into alternative venues known as "dark pools," which are less regulated, Schneiderman said.
(Read more: High-frequency trading: CNBC Explains)
He suggested reforms for stock exchanges, such as a proposal by University of Chicago economists that they process orders in batches rather than continuously, to ensure that price trumps technology in deciding who obtains a trade.
Schneiderman has been conducting a sweeping investigation of early access to data. Last month, Berkshire Hathaway's Business Wire said it would no longer sell potentially market-moving press releases directly to high frequency-trading companies after months of discussion with his office.
In January, BlackRock, the world's largest asset manager, agreed to end its analyst survey program worldwide, and 18 brokerages, including Goldman Sachs, JPMorgan Chase and Citigroup, later agreed to end their participation in such programs.
Last July, Thomson Reuters said it would suspend its early release of the widely watched Thomson Reuters/University of Michigan consumer sentiment data to a small group of clients in response to the probe.