Cyber Week in Review: February 15 , 2019

Here is a quick round-up of this week’s technology headlines and related stories you may have missed: 

Russia Unplugs Itself. Russia is planning to temporarily disconnect from the global internet at some point before April 1. The dramatic move is part of an effort to test and generate feedback for a proposed law, known as the Digital Economy National Program. Following years of trying to assert Russian independence in cyberspace, Russia is finally prepared to unplug. The country has already pioneered a BRICS-nations only DNS backup system, and the proposed law is another measure aimed at strengthening Russia’s defenses against a potential NATO cyberattack. The law would require that all domestic internet traffic travels through Russian-based servers approved by telecom regulator Roskomnazor, effectively cutting the country off from the global internet. For Moscow, there is much to like in a country-wide intranet: it would simplify censorship and prevent the interception of Russian internet data on foreign servers. If the technical test is successful, the Russian Parliament is likely to pass the bill in the near future.

Opening Pandora’s Box: OpenAI, a non-profit research institute, released a record-shattering unsupervised language model this week. The model was “trained simply to predict the next word in 40GB of Internet text." The results are remarkable: Give OpenAI’s system a writing prompt and it will write you a convincing essay, capturing many of the subtleties of human language that AI has long struggled to grasp. Yet the implications of such a breakthrough are cause for concern. OpenAI’s policy director Jack Clark compared the language-generating technology to the AI behind deepfakes, which commentators have also worried has the potential to create realistic-seeming false content. For this reason, OpenAI took the unusual step of refusing to publish the full model or its training data because of the potential for misuse. Such ethical concerns are increasing part and parcel with cutting-edge AI research. Last month, Google announced it would limit what the company published about its research, lest the technology falls into the wrong hands.

Trump AI policy: On Monday, the Trump administration released its “American AI Initiative” an executive order aiming to maintain American leadership in artificial intelligence. The order contains many of the same recommendations as an Obama-era 2016 report on AI: increasing AI research and development, building a larger AI workforce, and bolstering U.S. cyber defenses. Notably, the order does not allocate any new funding toward AI but rather directs agency chiefs to prioritize AI in fiscal year 2020 and beyond. The initiative also directs educational programs to “consider AI as a priority area” and to “give preference to American citizens” in grants, scholarships, military commissions, and course development. Furthermore, the executive order promises to protect Americans’ civil liberties and to increase access to federal government data. Critics were quick to note that the executive order does not mention immigration policy—considered a key source of AI talent—or address the workforce challenges caused by AI disruption.

Disinformation Moldova Style: Facebook announced this week that it has deleted 200 Facebook and Instagram accounts in response to election-based misinformation in Moldova. The small eastern European nation will hold parliamentary elections on February 24 under a new, controversial mixed electoral system. As one of the poorest countries in Europe, Moldova has been rated as one of the states in the region most vulnerable to misinformation campaigns. The accounts that Facebook removed had a combined following of more than 55,000 followers, a significant number given that only 1.6 million votes were cast in the last parliamentary election in 2014. The takedowns demonstrate the new, muscle response that Facebook is taking to ensure its platforms do not become weapons in electoral politics. Last week, WhatsApp—which Facebook owns—threatened to kick Indian political parties off WhatsApp if they were caught misusing the messaging service.