The first batch of reports assessing the so called “systemic risks” posed by the largest online platforms are in. These reports are a result of the Digital Services Act (DSA), Europe’s new law regulating platforms like Google, Meta, Amazon or X, and have been eagerly awaited by civil society groups across the globe. In their reports, companies are supposed to assess whether their services contribute to a wide range of barely defined risks. These go beyond the dissemination of illegal content and include vaguely defined categories such as negative effects on the integrity of elections, impediments to the exercise of fundamental rights or undermining of civic discourse. We have previously warned that the subjectivity of these categories invites a politization of the DSA.  

In view of a new DSA investigation into TikTok’s potential role in Romania’s presidential election, we take a look at the reports and the framework that has produced them to understand their value and limitations.  

A Short DSA Explainer  

The DSA covers a lot of different services. It regulates online markets like Amazon or Shein, social networks like Instagram and TikTok, search engines like Google and Bing, and even app stores like those run by Apple and Google. Different obligations apply to different services, depending on their type and size. Generally, the lower the degree of control a service provider has over content shared via its product, the fewer obligations it needs to comply with.   

For example, hosting services like cloud computing must provide points of contact for government authorities and users and basic transparency reporting. Online platforms, meaning any service that makes user generated content available to the public, must meet additional requirements like providing users with detailed information about content moderation decisions and the right to appeal. They must also comply with additional transparency obligations.  

While the DSA is a necessary update to the EU’s liability rules and improved users’ rights, we have plenty of concerns with the route that it takes:  

• We worry about the powers it gives to authorities to request user data and the obligation on providers to proactively share user data with law enforcement.  

• We are also concerned about the ways in which trusted flaggers could lead to the over-removal of speech, and  

• We caution against the misuse of the DSA’s mechanism to deal with emergencies like a pandemic. 

Introducing Systemic Risks 

The most stringent DSA obligations apply to large online platforms and search engines that have more than 45 million users in the EU. The European Commission has so far designated more than 20 services to constitute such “very large online platforms” (VLOPs) or “very large online search engines” (VLOSEs). These companies, which include X, TikTok, Amazon, Google Search, Maps and Play, YouTube and several porn platforms, must proactively assess and mitigate “systemic risks” related to the design, operation and use of their services. The DSA’s non-conclusive list of risks includes four broad categories: 1) the dissemination of illegal content, 2) negative effects on the exercise of fundamental rights, 3) threats to elections, civic discourse and public safety, and 4) negative effects and consequences in relation to gender-based violence, protection of minors and public health, and on a person’s physical and mental wellbeing.  

The DSA does not provide much guidance on how VLOPs and VLOSEs are supposed to analyze whether they contribute to the somewhat arbitrary seeming list of risks mentioned. Nor does the law offer clear definitions of how these risks should be understood, leading to concerns that they could be interpreted widely and lead to the extensive removal of lawful but awful content. There is equally little guidance on risk mitigation as the DSA merely names a few measures that platforms can choose to employ. Some of these recommendations are incredibly broad, such as adapting the design, features or functioning of a service, or “reinforcing internal processes”. Others, like introducing age verification measures, are much more specific but come with a host of issues and can undermine fundamental rights themselves.   

Risk Management Through the Lens of the Romanian Election 

Per the DSA, platforms must annually publish reports detailing how they have analyzed and managed risks. These reports are complemented by separate reports compiled by external auditors, tasked with assessing platforms’ compliance with their obligations to manage risks and other obligations put forward by the DSA.  

To better understand the merits and limitations of these reports, let’s examine the example of the recent Romanian election. In late November 2024, an ultranationalist and pro-Russian candidate, Calin Georgescu, unexpectedly won the first round of Romania’s presidential election. After reports by local civil society groups accusing TikTok of amplifying pro-Georgescu content, and a declassified brief published by Romania’s intelligence services that alleges cyberattacks and influence operations, the Romanian constitutional court annulled the results of the election. Shortly after, the European Commission opened formal proceedings against TikTok for insufficiently managing systemic risks related to the integrity of the Romanian election. Specifically, the Commission’s investigation focuses on “TikTok's recommender systems, notably the risks linked to the coordinated inauthentic manipulation or automated exploitation of the service and TikTok's policies on political advertisements and paid-for political content.” 

TikTok’s own risk assessment report dedicates eight pages to potential negative effects on elections and civic discourse. Curiously, TikTok’s definition of this particular category of risk focuses on the spread of election misinformation but makes no mention of coordinated inauthentic behavior or the manipulation of its recommender systems. This illustrates the wide margin on platforms to define systemic risks and implement their own mitigation strategies. Leaving it up to platforms to define relevant risks not only makes the comparison of approaches taken by different companies impossible, it can also lead to overly broad or narrow approaches—potentially undermining fundamental rights or running counter to the obligation to effectively deal with risks, as in this example. It should also be noted that mis- and disinformation are terms not defined by international human rights law and are therefore not well suited as a robust basis on which freedom of expression may be restricted.  

In its report, TikTok describes the measures taken to mitigate potential risks to elections and civic discourse. This overview broadly describes some election-specific interventions like labels for content that has not been fact checked but might contain misinformation, and describes TikTok’s policies like its ban of political ads, which is notoriously easy to circumvent. It does not entail any indication that the robustness and utility of the measures employed are documented or have been tested, nor any benchmarks of when TikTok considers a risk successfully mitigated. It does not, for example, contain figures on how many pieces of content receive certain labels, and how these influence users’ interactions with the content in question.  

Similarly, the report does not contain any data regarding the efficacy of TikTok’s enforcement of its political ads ban. TikTok’s “methodology” for risk assessments, also included in the report, does not help in answering any of these questions, either. And looking at the report compiled by the external auditor, in this case KPMG, we are once again left disappointed: KPMG concluded that it was impossible to assess TikTok’s systemic risk compliance because of two earlier, pending investigations by the European Commission due to potential non-compliance with the systemic risk mitigation obligations. 

Limitations of the DSA’s Risk Governance Approach 

What then, is the value of the risk and audit reports, published roughly a year after their finalization? The answer may be very little.  

As explained above, companies have a lot of flexibility in how to assess and deal with risks. On the one hand, some degree of flexibility is necessary: every VLOP and VLOSE differs significantly in terms of product logics, policies, user base and design choices. On the other hand, the high degree of flexibility in determining what exactly a systemic risk is can lead to significant inconsistencies and render risk analysis unreliable. It also allows regulators to put forward their own definitions, thereby potentially expanding risk categories as they see fit to deal with emerging or politically salient issues.  

Rather than making sense of diverse and possibly conflicting definitions of risks, companies and regulators should put forward joint benchmarks, and include civil society experts in the process. 

Speaking of benchmarks: There is a critical lack of standardized processes, assessment methodologies and reporting templates. Most assessment reports contain very little information on how the actual assessments are carried out, and the auditors’ reports distinguish themselves through an almost complete lack of insight into the auditing process itself. This information is crucial, but it is near impossible to adequately scrutinize the reports themselves without understanding whether auditors were provided the necessary information, whether they ran into any roadblocks looking at specific issues, and how evidence was produced and documented. And without methodologies that are applicable across the board it will remain very challenging, if not impossible, to compare approaches taken by different companies.  

The TikTok example shows that the risk and audit reports do not contain the “smoking gun” some might have hoped for. Besides the shortcomings explained above, this is due to the inherent limitations of the DSA itself. Although the DSA attempts to take a holistic approach to complex societal risks that cut across different but interconnected challenges, its reporting system is forced to only consider the obligations put forward by the DSA itself. Any legal assessment framework will struggle to capture complex societal challenges like the integrity of elections or public safety. In addition, phenomena as complex as electoral processes and civic discourse are shaped by a range of different legal instruments, including European rules on political ads, data protection, cybersecurity and media pluralism, not to mention countless national laws. Expecting a definitive answer on the potential implications of large online services on complex societal processes from a risk report will therefore always fall short.  

The Way Forward  

The reports do present a slight improvement in terms of companies’ accountability and transparency. Even if the reports may not include the hard evidence of non-compliance some might have expected, they are a starting point to understanding how platforms attempt to grapple with complex issues taking place on their services. As such, they are, at best, the basis for an iterative approach to compliance. But many of the risks described by the DSA as systemic and their relationships with online services are still poorly understood.  

Instead of relying on platforms or regulators to define how risks should be conceptualized and mitigated, a joint approach is needed—one that builds on expertise by civil society, academics and activists, and emphasizes best practices. A collaborative approach would help make sense of these complex challenges and how they can be addressed in ways that strengthen users’ rights and protect fundamental rights.  

California’s 2024 legislative session has officially adjourned, and it’s time to reflect on the wins and losses that have shaped Californians’ digital rights landscape this year.

EFF monitored nearly 100 bills in the state this session alone, addressing a broad range of issues related to privacy, free speech, and innovation. These include proposed standards for Artificial Intelligence (AI) systems used by state agencies, the intersection of AI and copyright, police surveillance practices, and various privacy concerns. While we have seen some significant victories, there are also alarming developments that raise concerns about the future of privacy protection in the state.

Celebrating Our Victories

This legislative session brought some wins for privacy advocates—most notably the defeat of four dangerous bills: A.B. 3080, A.B. 1814, S.B. 1076, and S.B. 1047. These bills posed serious threats to consumer privacy and would have undermined the progress we’ve made in previous years.

First, we commend the California Legislature for not advancing A.B. 3080, “The Parent’s Accountability and Child Protection Act” authored by Assemblymember Juan Alanis (Modesto). The bill would have created powerful incentives for “pornographic internet websites” to use age-verification mechanisms. The bill was not clear on what counts as “sexually explicit content.” Without clear guidelines, this bill will further harm the ability of all youth—particularly LGBTQ+ youth—to access legitimate content online. Different versions of bills requiring age verification have appeared in more than a dozen states. We understand Asm. Alanis' concerns, but A.B. 3080 would have required broad, privacy-invasive data collection from internet users of all ages. We are grateful that it did not make it to the finish line.

Second, EFF worked with dozens of organizations to defeat A.B. 1814, a facial recognition bill authored by Assemblymember Phil Ting (San Francisco). The bill attempted to expand the use of facial recognition software by police to “match” images from surveillance databases to possible suspects. Those images could then be used to issue arrest warrants or search warrants. The bill merely said that these matches can't be the sole reason for a warrant to be issued—a standard that has already failed to stop false arrests in other states.  Police departments and facial recognition companies alike both currently maintain that police cannot justify an arrest using only algorithmic matches–so what would this bill really change? The bill only gave the appearance of doing something to address face recognition technology's harms, while allowing the practice to continue. California should not give law enforcement the green light to mine databases, particularly those where people contributed information without knowledge that it would be accessed by law enforcement. You can read more about this bill here, and we are glad to see the California legislature reject this dangerous bill.

EFF also worked to oppose and defeat S.B. 1076, by Senator Scott Wilk (Lancaster). This bill would have weakened the California Delete Act (S.B. 362). Enacted last year, the Delete Act provides consumers with an easy “one-click” button to request the removal of their personal information held by data brokers registered in California. By January 1, 2026. S.B. 1076 would have opened loopholes for data brokers to duck compliance. This would have hurt consumer rights and undone oversight on an opaque ecosystem of entities that collect then sell personal information they’ve amassed on individuals. S.B. 1076 would have likely created significant confusion with the development, implementation, and long-term usability of the delete mechanism established in the California Delete Act, particularly as the California Privacy Protection Agency works on regulations for it. 

Lastly, EFF opposed S.B. 1047, the “Safe and Secure Innovation for Frontier Artificial Intelligence Models Act” authored by Senator Scott Wiener (San Francisco). This bill aimed to regulate AI models that might have "catastrophic" effects, such as attacks on critical infrastructure. Ultimately, we believe focusing on speculative, long-term, catastrophic outcomes from AI (like machines going rogue and taking over the world) pulls attention away from AI-enabled harms that are directly before us. EFF supported parts of the bill, like the creation of a public cloud-computing cluster (CalCompute). However, we also had concerns from the beginning that the bill set an abstract and confusing set of regulations for those developing AI systems and was built on a shaky self-certification mechanism. Those concerns remained about the final version of the bill, as it passed the legislature.

Governor Newsom vetoed S.B. 1047; we encourage lawmakers concerned about the threats unchecked AI may pose to instead consider regulation that focuses on real-world harms.  

Of course, this session wasn’t all sunshine and rainbows, and we had some big setbacks. Here are a few:

The Lost Promise of A.B. 3048

Throughout this session, EFF and our partners supported A.B. 3048, common-sense legislation that would have required browsers to let consumers exercise their protections under the California Consumer Privacy Act (CCPA). California is currently one of approximately a dozen states requiring businesses to honor consumer privacy requests made through opt–out preference signals in their browsers and devices. Yet large companies have often made it difficult for consumers to exercise those rights on their own. The bill would have properly balanced providing consumers with ways to exercise their privacy rights without creating burdensome requirements for developers or hindering innovation.

Unfortunately, Governor Newsom chose to veto A.B. 3048. His veto letter cited the lack of support from mobile operators, arguing that because “No major mobile OS incorporates an option for an opt-out signal,” it is “best if design questions are first addressed by developers, rather than by regulators.” EFF believes technologists should be involved in the regulatory process and hopes to assist in that process. But Governor Newsom is wrong: we cannot wait for industry players to voluntarily support regulations that protect consumers. Proactive measures are essential to safeguard privacy rights.

This bill would have moved California in the right direction, making California the first state to require browsers to offer consumers the ability to exercise their rights. 

Wrong Solutions to Real Problems

A big theme we saw this legislative session were proposals that claimed to address real problems but would have been ineffective or failed to respect privacy. These included bills intended to address young people’s safety online and deepfakes in elections.

While we defeated many misguided bills that were introduced to address young people’s access to the internet, S.B. 976, authored by Senator Nancy Skinner (Oakland), received Governor Newsom’s signature and takes effect on January 1, 2027. This proposal aims to regulate the "addictive" features of social media companies, but instead compromises the privacy of consumers in the state. The bill is also likely preempted by federal law and raises considerable First Amendment and privacy concerns. S.B. 976 is unlikely to protect children online, and will instead harm all online speakers by burdening free speech and diminishing online privacy by incentivizing companies to collect more personal information.

It is no secret that deepfakes can be incredibly convincing, and that can have scary consequences, especially during an election year. Two bills that attempted to address this issue are A.B. 2655 and A.B. 2839. Authored by Assemblymember Marc Berman (Palo Alto), A.B. 2655 requires online platforms to develop and implement procedures to block and take down, as well as separately label, digitally manipulated content about candidates and other elections-related subjects that creates a false portrayal about those subjects. We believe A.B. 2655 likely violates the First Amendment and will lead to over-censorship of online speech. The bill is also preempted by Section 230, a federal law that provides partial immunity to online intermediaries for causes of action based on the user-generated content published on their platforms. 

Similarly, A.B. 2839, authored by Assemblymember Gail Pellerin (Santa Cruz), not only bans the distribution of materially deceptive or altered election-related content, but also burdens mere distributors (internet websites, newspapers, etc.) who are unconnected to the creation of the content—regardless of whether they know of the prohibited manipulation. By extending beyond the direct publishers and toward republishers, A.B. 2839 burdens and holds liable republishers of content in a manner that has been found unconstitutional.

There are ways to address the harms of deepfakes without stifling innovation and free speech. We recognize the complex issues raised by potentially harmful, artificially generated election content. But A.B. 2655 and A.B. 2839, as written and passed, likely violate the First Amendment and run afoul of federal law. In fact, less than a month after they were signed, a federal judge put A.B. 2839’s enforcement on pause (via a preliminary injunction) on First Amendment grounds.

Privacy Risks in State Databases

We also saw a troubling trend in the legislature this year that we will be making a priority as we look to 2025. Several bills emerged this session that, in different ways, threatened to weaken privacy protections within state databases. Specifically,  A.B. 518 and A.B. 2723, which received Governor Newsom’s signature, are a step backward for data privacy.

A.B. 518 authorizes numerous agencies in California to share, without restriction or consent, personal information with the state Department of Social Services (DSS), exempting this sharing from all state privacy laws. This includes county-level agencies, and people whose information is shared would have no way of knowing or opting out. A. B. 518 is incredibly broad, allowing the sharing of health information, immigration status, education records, employment records, tax records, utility information, children’s information, and even sealed juvenile records—with no requirement that DSS keep this personal information confidential, and no restrictions on what DSS can do with the information.

On the other hand, A.B. 2723 assigns a governing board to the new “Cradle to Career (CTC)” longitudinal education database intended to synthesize student information collected from across the state to enable comprehensive research and analysis. Parents and children provide this information to their schools, but this project means that their information will be used in ways they never expected or consented to. Even worse, as written, this project would be exempt from the following privacy safeguards of the Information Practices Act of 1977 (IPA), which, with respect to state agencies, would otherwise guarantee California parents and students:

1.     the right for subjects whose information is kept in the data system to receive notice their data is in the system;
2.     the right to consent or, more meaningfully, to withhold consent;
3.     and the right to request correction of erroneous information.

By signing A.B. 2723, Gov. Newsom stripped California parents and students of the rights to even know that this is happening, or agree to this data processing in the first place. 

Moreover, while both of these bills allowed state agencies to trample on Californians’ IPA rights, those IPA rights do not even apply to the county-level agencies affected by A.B. 518 or the local public schools and school districts affected by A.B. 2723—pointing to the need for more guardrails around unfettered data sharing on the local level.

A Call for Comprehensive Local Protections

A.B. 2723 and A.B. 518 reveal a crucial missing piece in Californians' privacy rights: that the privacy rights guaranteed to individuals through California's IPA do not protect them from the ways local agencies collect, share, and process data. The absence of robust privacy protections at the local government level is an ongoing issue that must be addressed.

Now is the time to push for stronger privacy protections, hold our lawmakers accountable, and ensure that California remains a leader in the fight for digital privacy. As always, we want to acknowledge how much your support has helped our advocacy in California this year. Your voices are invaluable, and they truly make a difference.

Let’s not settle for half-measures or weak solutions. Our privacy is worth the fight.

This post was written by EFF intern Nazli Ungan as an update to a 2020 Deeplinks post by Cindy Cohn.

Everyone wants an election that is secure and reliable and that will ensure that the voters’ actual choices are reflected in the results. That’s as true as we head into the 2024 U.S. general elections as it always has been.

At the same time, not every problem in voting technology or systems is worth pulling the fire alarm—we have to look at the bigger story and context. And we have to stand down when our worst fears turn out to be unfounded.

Resilience is the key word when it comes to the security and the integrity of our elections. We need our election systems to be technically and procedurally resilient against potential attacks or errors. But equally important, we need the voting public to be resilient against false or unfounded claims of attack or error. Luckily, our past experiences and the work of election security experts have taught us a few lessons on when to worry and when to not.

See EFF's handout on Election Security here: https://www.eff.org/document/election-security-recommendations

We Need Risk-Limiting Audits

First, and most importantly, it is critical to have systems in place to support election technology and the election officials who run it. Machines may fail, humans may make errors. We cannot simply assume that there will not be any issues in voting and tabulation. Instead, there must be built-in safety measures that would catch any issues that may affect the official election results.  

It is critical to have systems in place to support election technology and the election officials who run it.

The most important of these is performing routine, post-election Risk-Limiting Audits after every election. RLAs should occur even if there is no apparent reason to suspect the accuracy of the results. Risk-limiting audits are considered the gold standard of post-election audits and they give the public justified confidence in the results. This type of audit entails manually checking randomly selected ballots until there is convincing evidence that the election outcome is correct. In many cases, it can be performed by counting only a small fraction of ballots cast making it cheap enough to be performed in every election. When the margins are tighter, a greater fraction of the votes are required to be hand counted, but this is a good thing because we want to scrutinize close contests more strictly to make sure the right person won the race. Some states have started requiring risk-limiting audits and the rest should catch up!

 We (and many others in the election integrity community) also continue to push for more transparency in election systems, more independent testing and red-team style attacks, including end-to-end pre-election testing.

And We Need A Paper Trail

Second, voting on paper ballots continues to be extremely important and the most secure strategy. Ideally, all voters should use paper ballots marked by hand, or with an assistive device, and verify their votes before casting. If there is no paper record, there is no way to perform a post-election audit, or recount votes in the event of an error or a security incident. On the other hand, if voters vote on paper, they can verify their choices are recorded accurately. More importantly, election officials can hand count a portion of the paper ballots to make sure they match with the electronic vote totals and confirm the accuracy of the election results. 

What happened in Antrim County, Michigan in the 2020 general elections illustrates the importance of paper ballots. Immediately after the 2020 elections, Antrim County published inaccurate unofficial results, and then restated these results three times to correct the errors, which led to conspiracy theories about the voting systems used there. Fortunately, Antrim County voters had voted on paper ballots, so Michigan was able to confirm the final presidential results by conducting a county-wide hand count and affirm them by a state-wide risk-limiting audit pilot. This would not have been possible without paper ballots.  

And we can’t stop there, because not every paper record is created equal. Some direct recording electronic systems are equipped with a type of Voter-Verified Paper Audit Trail that make it difficult for voters to verify their selections and for election officials to use in audits and recounts. The best practice is to have all votes cast on pre-printed paper ballots, marked by hand or an assistive ballot marking device.  

Third, it is important to have the entire voting technical system under the control of election officials so that they can investigate any potential problems, which is one of the reasons why internet voting remains a bad, bad idea. There are “significant security, privacy, and ballot secrecy challenges” associated with electronic ballot return systems and they make it “possible for a single attacker to alter thousands or even millions of votes.” Maybe in the future we will have tools to limit the risks of internet voting. But until then, we should reject any proposal that includes electronic ballot return over the internet. Speaking about the internet, voting machines should never connect to the internet, dial a modem, or communicate wirelessly. 

Internet voting remains a bad, bad idea

Fourth, every part of the voting process that relies on technology must have paper backups so that voting can continue even when the machines fail. This includes paper backups for electronic pollbooks, emergency paper ballots in case voting machines fail, and provisional ballots in case there voter eligibility cannot be confirmed. 

Stay Vigilant and Informed

Fifth, we should continue to be vigilant. Election officials have come a long way from when we started raising concerns about electronic voting machines and systems. But the public should keep watching and, when warranted, not be afraid to raise or flag things that seem strange. For example, if you see something like voting machines “flipping” the votes, you should tell the poll workers. This doesn’t necessarily mean there has been a security breach; it can be as simple as a calibration error, but it can mean lost votes. Poll workers can and should address the issue immediately by providing voters with emergency paper ballots. 

Sixth, not everything that seems out of the ordinary may be reason to worry. We should build societal resistance to disinformation. CISA's Election Security Rumor vs. Reality website is a good resource that addresses election security rumors and educates us on when we need to be or don’t need to be alarmed. State-specific information is also available online. If we see or hear anything odd about what is happening at a particular locality, we should first hear what the election officials on the ground have to say about it. After all, they were there! We should also pay attention to what non-partisan election protection organizations, such as Verified Voting, say about the incident.  

The 2024 presidential election is fast approaching and there may be many claims of computer glitches and other forms of manipulation concerning our voting systems in November. Knowing when to worry and when NOT to worry will continue to be extremely important.  

In the meantime, the work of securing our elections and building resilience must continue. While not every glitch is worrisome, we should not dismiss legitimate security concerns. As often said: election security is a race without a finish line!