A Twitter security flaw is being widely exploited on Twitter, showing remote content from third-party websites without user’s consent.
The flaw uses a JavaScript function called onMouseOver() which creates an event when the mouse is passed over a text or link. Any user can use this flaw to
create simple popups, redirect the page to somewhere else, retweet some messages or hide parts of the message.
The link has to be constructed in such a way that it starts with http://twitter.com/<text>@”onmouseover=”<code>”
The problem is that Twitter doesn’t filter the code, it instead executes it.
To overcome the problem, use some 3rd party websites to work with Twitter. These website use the API directly and not the twitter.com website GUI.
Another solution is to use the mobile website, which doesn’t seem to have the flaw.
Update: Twitter fixed the problem : http://blog.twitter.com/2010/09/all-about-onmouseover-incident.html