Skip to content Skip to navigation

Friday Cyber News, May 13 2016

Cyber technology-related news and links from around the web, for the week of 5/7 - 5/13:

1. The Turing Test posits that an AI indistinguishable from a human can be identified by having the equivalent of a Gchat conversation with it. Another way of re-phrasing that test is, can we design an AI we would want to flirt with? One step is to teach a bot to become proficient at understanding human language, including speech. That helps us do everything from provide better customer service to do legal work to be a teaching assistant to spot fake Amazon reviews. But flirt with enough people, call enough customer support representatives, and interact with sufficiently many lawyers, and you'll find some bad apples: to hasten the inevitable, researchers have published a guide to building a malevolent AI. [New Republic; CNET; TechCrunch; Quartz; The Next Web; Wirecutter; Arxiv]

2. Facebook's process of selecting what news makes it into their trending topics list is subject to human--and possibly anti-conservative--bias. The New Yorker delves deeper, explaining why human curators are necessary ("cats" would otherwise be perpetually trending) and why it matters that Facebook is biased, when we know that other sources of news have a slant as well (the Economist and Fox vs. MSNBC and the New Yorker itself, for example). [Guardian]

3. Oregonian Senator Wyden claims that the DOJ lied in a 2003 Office of Legal Counsel opinion on telecommunications companies providing data (supposedly; the opinion has not yet been made public but is thought to govern when the government can access people's data). Deirdre Walsh at the ODNI struck back at the Senator's pro-privacy activities in a roundabout way by sending an open letter to Senator Wyden arguing that the Berkman Center's "Don't Panic" report, which details law enforcement's tools to prevent going dark, draws inaccurate conclusions. Susan Landau offers a counterpoint, and the director of Europol says that encryption is a problem in 75% of their cases, though that seems high. (So, do panic?) [The Hill;; Harvard; Lawfare; Twitter]

4. Related to the previous, are US courts going dark? And how we can recognize poorly thought-out "exceptional access" provisions, from the Keys Under Doormats team. [Just Security; Lawfare]

5. Research into the human side of cyber crime--how users pick passwords, respond to social engineering, and internalize security warnings--is picking up with the realization that to outsmart hackers, we need to better understand their motivations, and our vulnerabilities. [Nature]

6. Is "sextortion"--the stealing or coerced provision of sexually explicit data, such as images and video--cyber bullying, data theft, or sexual assault? Or all three? A new report from the Brookings institution looks at the legal status of online sex crime. [NY Times]

7. Hacking of the SWIFT payment-processing network, implicated in the $81M Bangladeshi bank theft, is now being investigated as an inside job. Those affected (SWIFT, the Bangladeshi bank, and the Federal Reserve Bank of New York) are joining forces to recover their lost assets. Meanwhile, a Vietnamese bank was hacked using similar methods. [WSJ; CNBC; Financial Times]

8. The FCC and FTC are investigating how mobile phone carriers and manufacturers handle security updates, to improve the process for patching vulnerabilities, particularly for the distributed Android ecosystem. [Reuters]

9. Twitter barred intelligence agencies from accessing analysis of tweets provided by a third party, apparently realizing after two years that its own privacy policies ban the sale of its data to the government for surveillance. [WSJ]

10. Rich people have access to the internet. Poor people don't. [Public Integrity]

Thanks for reading,

Stanford Cyber Initiative

(To suggest an item for this list, please email This email contains links that may not work if HTML is not supported in your mailbox. You can view news from past weeks, subscribe, and unsubscribe at