Group items tagged

While AI tools still exist in a relative legal vacuum, this blog post explores: 1) the extent of protection granted to algorithms as trade secrets with exceptions of overriding public interest; 2) how the new generation of regulations on the EU and national levels attempt to provide algorithm transparency while preserving trade secrecy; and 3) why the latter development is not a futile endeavour. 

most complex algorithms dominating our lives (including those developed by Google and Facebook), are proprietary, i.e. shielded as trade secrets, while only a negligible minority of algorithms are open source. 

Article 2 of the EU Trade Secrets Directive

elf-assessment reports submitted by Facebook, Google, Microsoft, Mozilla and Twitter

bserved that “[a]ll platform signatories deployed policies and systems to ensure transparency around political advertising, including a requirement that all political ads be clearly labelled as sponsored content and include a ‘paid for by’ disclaimer.”

While some of the platforms have gone to the extent of banning political ads, the transparency of issue-based advertising is still significantly neglected.

New responsibilities arise from these changes.

Greater transparency will disclose whether further regulation is required and make it better targeted, providing specific remedies for clearly identified ills.

If Facebook and others must account in detail to an electoral commission or data protection authority for micro-targeting or “dark” ads, are forbidden from deleting certain relevant data, and must submit to algorithm audits, they will forced to foresee and to try to solve some of the problems which they have been addressing so slowly

Procedural accountability” was a focus of discussion at the March 2018 workshop on platform responsibility convened by LSE’s Truth, Trust and Technology Commission. The idea is that firms should be held to account for the effectiveness of their internal processes in tackling the negative social impact of their services.

o be credible and trusted, information disclosed by online firms will need to be independently verified.

Piloting a Transparency Reporting Framework

In a six-month period from October 2017 to March 20178, 21m sexually explicit pictures, 3.5m graphically violent posts and 2.5m forms of hate speech were removed from its site. These figures help reveal some striking points.

As expected, the data indicates that the problem is getting worse.

For instance, between January and March it was estimated that for every 10,000 messages online, between 22 and 27 contained graphic violence, up from 16 to 19 in the previous three months.

he shift from product- and service-based to platform-based business creates a new set of platform governance implications — especially when these businesses rely upon shared infrastructure from a small, powerful group of technology providers (Figure 1).

The industries in which AI is deployed, and the primary use cases it serves, will naturally determine the types and degrees of risk, from health and physical safety to discrimination and human-rights violations. Just as disinformation and hate speech are known risks of social media platforms, fatal accidents are a known risk of automobiles and heavy machinery, whether they are operated by people or by machines. Bias and discrimination are potential risks of any automated system, but they are amplified and pronounced in technologies that learn, whether autonomously or by training, from existing data.

Business Model-Specific Implications

his set of rules creates a novel regulatory model

Again, leaving regulatory powers to a private entity without any public oversight is clearly not the right solution. But this is also not what, in my opinion, the new AVMSD does

But without transparency and information about individual cases, you surely can’t say whether the takedowns are really improving the media environment, or the providers are just trying to get rid of any controversial content – or, indeed, the content somebody just happens to be complaining about.

We often talk about the “online environment.” This metaphorical language makes it seem like the online space looks similar to our offline world. For example, the term “information pollution,” coined by Claire Wardle, is increasingly being used to discuss disinformation online.  

It is even harder to prove direct connections between online platforms and offline harms. This is partly because platforms are not transparent.

Finally, this analogy reminds us that both problems are dispiritingly hard to solve. Two scholars, Whitney Phillips and Ryan Milner, have suggested that our online information problems are ecosystemic, similar to the climate crisis.

. In e-commerce in particular, market surveillance authorities have great difficulty tracing non-compliant products imported into the Union and identifying the responsible entity within their jurisdiction.

In its 2017 work programme 4 , the Commission announced an initiative to strengthen product compliance and enforcement Union harmonisation legislation on products, as part of the 'Goods Package'. The initiative is to address the increasing amount of non-compliant products on the Union market while offering incentives to boost regulatory compliance and ensuring fair and equal treatment that will benefit of businesses and citizens.

The development of e-commerce is also due to a great extent to the proliferation of information society service providers, normally through platforms and for remuneration, which offer intermediary services by storing third party content, but without exercising any control over such content, thus not acting on behalf of an economic operator. Removal of content regarding non-compliant products or where it is not feasible blocking access to non-compliant products offered through their services should be without prejudice to the rules laid down in Directive 2000/31/EC of the European Parliament and of the Council 55 . In particular, no general obligation should be imposed on service providers to monitor the information which they transmit or store, nor should a general obligation be imposed upon them to actively seek facts or circumstances indicating illegal activity. Furthermore, hosting service providers should not be held liable as long as they do not have actual knowledge of illegal activity or information and are not aware of the facts or circumstances from which the illegal activity or information is apparent.

On 19 September, Google and the Association to Combat Audiovisual Piracy (Association de Lutte contre la Piraterie Audiovisuelle - “ALPA”) signed a partnership agreement aimed at effectively reinforcing copyright protection for the on-line exploitation of audiovisual works.

under the auspices of the National Centre for the Cnema (Centre National du Cinéma - “the CNC”

The EU’s overall objectives are certainly well-intended. However, many concerns remain, for instance:

The DSA should tackle bad players and behaviours regardless of the platform’s size and country of origin. Having a specific regime for “very large online platforms” with additional obligations leaves the door open for rogue players to simply move to smaller digital service providers that are subject to a lighter regime.

To prevent legal uncertainty, the DSA should have a clear scope focusing on illegal content, products and services. The rules should be horizontal and principle-based, and could in a second phase be complemented with more targeted measures (legislative and non-legislative) to tackle specific concerns. 

The explicit reference to human rights in its charter acknowledges that companies have a role in protecting and enforcing human rights.

This is consistent with efforts by the United Nations and other advocacy efforts to create standards on how businesses should be held accountable for human rights abuses. In light of Facebook’s entanglement in misinformation, scandals and election falsehoods, as well as genocide and incitement of violence, it seems particularly pertinent for the company.

To date, we have assigned such decision-making powers to states, many of which are accountable to their citizens. Facebook, on the other hand, is unaccountable to citizens in nations around the world, and a single individual (Mark Zuckerberg) holds majority decision-making power at the company.

n the up-side, flexible standards would give platforms more leeway to figure out meaningful technical improvements, and perhaps arrive at more nuanced automated assessment of content over tim

The down-sides of open-ended SDOC standards could be considerable, though. Proactive measures devised by platforms themselves would, even when coupled with transparency obligations, be far less subject to meaningful public review, accountability,