Google Allo attempts to meet privacy advocates and the law halfway. But is the biggest hurdle still the interface itself?
Google Allo, a “smart reply” IM bot released on Wednesday, incorporates a number of existing Google technologies that furthers the company’s application of artificial intelligence to its products. It also continues the long-running battle between privacy advocates and anticipated obligations from law enforcement agencies about encryption.
Google first announced the Allo bot at its annual 2016 I/O developer conference. Prior to its release, they already clarified that the end-to-end user encryption technology would be available as an option in incognito mode, not as a default like it is for its competitors WhatsApp, Viber, and iMessage. But with its release, it is also clear that all messages not sent in incognito mode will be stored, rather than automatically deleted:
— Edward Snowden (@Snowden) September 21, 2016
Meeting the FBI in the middle
Initially, Google security engineers envisioned that Allo would “with one touch” let the user “always chat in incognito mode going forward” so that messages would be auto-deleted and end-to-end encrypted. This would have put it in the same league as WhatsApp or Snapchat, but was ultimately not adopted.
The decision to allow message storage as the default setting is being billed as part of the company’s own user data-mining strategy, but also has larger legal implications.
By going about the privacy settings in this manner, Google Allo is closer in line with the hopes of law enforcement and intelligence agencies like the FBI as these products come to market. The agency regards auto-delete messaging and end-to-end encryption as impediments to criminal investigations and has confronted Apple and other companies over it in light of recent incidents of domestic terrorism.
Smart memes but reluctant users
Privacy, though, was only part of the pitch. Allo is part of what Google CEO Sundar Pichai described to Forbes as the company’s long-term approach to AI, in which these machine learning systems eventually become part of most devices and therefore part of “your day, and in an ambient way, [as] things are there to help” customers.
Google has heavily advertised that Allo is capable of machine learning, like its new Inbox app, to automate responses as “smart replies” based on user behavior. It is also, therefore, another way of bringing Google Assistant to Android and iOS users.
Google Assistant is further configured to supply what it believes to be relevant text and media responses to comments: basically, it’s a meme database, albeit one apparently dialed down to an E for Everyone rating, as reddit Android users have found that it doesn’t recognize curse words. Early reviews suggest that the feature is still a bit stilted even if it’s a major step up from first generation “smart” assistants like Microsoft Agent.
Currently, there is no desktop or Web version, which as Dow Jones noted in May, puts it at a disadvantage against Facebook Messenger or WhatsApp. But privacy disputes, cross-platform access, and machine learning hiccups aside the biggest challenge facing Allo and all such bots is the default human setting where “the presence of an anthropomorphized helper can undermine individuals’ perceived autonomy.”