Artificial Intelligence versus Human Nature: Protecting Ourselves from the Perils of DAO-based Collective Investment Schemes


Dan Awrey
Professor of Law at Cornell Law School


Time to read

5 Minutes

If you believe the Fintech Intelligentsia, there is a revolution coming to the financial services industry.  This revolution will not be led by PayPal, or Alibaba, or SoFi – or any of the other darlings of the emerging fintech sector.  The vanguard of this revolution will instead be a rather benign sounding innovation known as a ‘decentralized autonomous organization’ (DAO).  In a nutshell, a DAO is a collection of individuals whose relationships with one another are governed by rules-based computer protocols known as ‘smart contracts’.  These smart contracts use modus ponens (if/then) logic to create what are in theory self-executing, self-enforcing state-contingent contracts.  Thus, for example, a group of individuals could write a smart contract stipulating that if this blog post is viewed by more than one hundred people, then the author would automatically receive some modest payment.  These smart contracts then send instructions to blockchain or other distributed ledger platforms: theoretically enabling the immediate transfer, verification and recordkeeping of payments in the form of Bitcoin, Ether or other crypto-currencies.  No lawyers.  No courts.  No pesky government oversight.  No problem?

One of the frequently cited potential uses of DAOs is as a technological platform for collective investment schemes. DAO-based collective investment schemes enable investors to pool capital and then collectively identify the projects or assets into which that capital should be invested.  This business model thus closely resembles that of a conventional collective investment scheme, with the crucial difference that – instead of contracting with a third party portfolio manager – the investors in a DAO make investment decisions on the basis of a smart contract reflecting a pre-determined decision-making (e.g. majority) rule.  This smart contract also stipulates other important terms such as when and how an investor is entitled to liquidate their investment.

At first glance, the absence of a third party portfolio manager might appear to ameliorate many of the agency cost problems that motivate public regulatory regimes such as the EU Alternative Fund Managers and UCITS Directives that impose investor protection, microprudential, conduct of business, and other requirements on collective investment schemes.  In theory, investors are also free to liquidate their investment in a DAO if they do not agree with a collective investment decision.  For many, these characteristics justify a strict caveat emptor approach toward both the private enforcement of the underlying smart contracts (along with the natural language contracts which often accompany them) and the public regulation of DAO-based collective investment schemes.

Upon closer inspection, however, a caveat emptor approach may not be in either the short or longer term interests of DAO-based collective investment schemes or their investors.   First, financial contracts are complex.  The shift from natural language contractual terms to rules embedded within hundreds, if not thousands, of lines of computer code – or, indeed, a combination of the two – seems unlikely to reduce this complexity.  Along the same vein, embedding contractual payoffs in complex computer code gives investors with a greater degree of technological sophistication an inherent advantage over other investors.  The resulting asymmetries of information and expertise will be particularly acute where the individuals writing the code invest alongside other, less informed, investors.  These asymmetries of information and expertise may also result in a form of ‘shrouded’ pricing, with technologically sophisticated investors able to systemically exploit their superior knowledge of the relevant smart contracts in order to extract value from less informed – or myopic – investors.  Programmers may also write smart contracts that seek to exploit the heuristics and biases of these less informed investors.  Where they exist, these problems will undermine the effective market discipline needed to make caveat emptor anything more than simply a license to rip off less sophisticated investors.

The prospect of such exploitation has been driven home by the recent ‘hack’ of between $50-60 million from a DAO-based collective investment scheme known as ‘the DAO’ (see Bloomberg, June 17 2016).  The DAO hack is interesting for two reasons.  First, within its initial 28-day funding window, the DAO raised over $150 million from more than 11,000 individual investors (see CoinDesk, June 25 2016).  Some media reports referred to it as the largest crowd funding in history (ibid.).  Second, the hackers were able to siphon value from other investors within the parameters of the rules laid down in the DAO’s smart contracts (if not necessarily the natural language sales material which the investors may have relied upon).  In effect, the hackers were able to use their technical knowledge to spot ‘bugs’ embedded in the code and exploit them for their personal gain.  Caveat emptor indeed.

Second, even the smartest smart contract is invariably incomplete.  This was evident from the DAO hack – where the speed and effectiveness of the DAO’s response was arguably undermined by the necessity of playing by its own rules relating to collective decision-making.  Liquidity management provides another illustrative example.  The speed and ease with which investors will be able to liquidate their investment in a DAO-based collective investment scheme will be a function of, inter alia, prevailing market conditions, the liquidity profile of the underlying assets, and the number of other investors who wish to liquidate their investments at the same time.  As any human portfolio manager will tell you, managing the liquidity of a collective investment scheme – i.e. knowing precisely when to sell assets in order to fund redemptions while maximizing value for investors – is a craft skill honed over many years of experience.  Writing a complete state-contingent contract that fully parameterizes each of the relevant variables and produces an executable instruction – sell/hold – for each potential future state of the world would thus be extremely, if not prohibitively, costly.

There exists no shortage of potential ways that this contractual incompleteness might result in the destruction of investor wealth.  What if a DAO executes an immediate sale of an asset into an illiquid market where waiting might have yielded a higher price and delivered more value for investors?  What if sell orders by multiple DAO-based collective investment schemes operating on the basis of similar underlying smart contracts create a self-reinforcing negative feedback loop, driving down prices, and destroying value?  And what if investors, perhaps anticipating this negative feedback loop, all simultaneously rush for the exit?  To observe the possible unintended consequences, one need look no further than the role of high frequency trading – itself a riff on smart contract technology – in precipitating the ‘Flash Crash’ of May 2010.

So if caveat emptor is not the right approach, what is the best way forward?  One approach of course is to wait for courts to fill in the contractual gaps and identify the legal grounds, if any, for holding individuals like the DAO hackers liable for their actions.  Indeed, lawyers in several jurisdictions are already burning the midnight oil developing theories of liability for the DAO hack – from civil and criminal liability, to tort and even constructive trusts and implied partnerships (see CoinDesk, June 17 2016).  Inevitably, however, this will be a slow, uneven process.  It is also unclear whether any of these avenues of recourse will bear remedial fruit.  More broadly, judges – focusing rightly as they do on the idiosyncratic facts of the case before them – may render decisions that run counter to the expectations of investors or, in extremis, undermine the institutional viability of the DAO model.  In the interim, DAO investors will have to live with a high degree of uncertainty surrounding the level of investor protection associated with DAO-based collective investment schemes.

Another approach is to bring DAOs within the perimeter of the public regulatory regimes designed to protect investors in collective investment schemes.  What precisely this regulation would look like would of course require careful thought.  Old approaches to regulating collective investment schemes may not be appropriate for the new world of DAOs.  There is also the risk that public regulatory intervention would preempt the emergence and evolution of private ordering designed to better protect investors.  It may also stifle welfare enhancing innovation, or simply drive this innovation into jurisdictions with less onerous regulatory regimes.  Addressing these challenges will be no easy task.  Over the longer term, however, addressing these challenges may be precisely what is needed if we are to engender confidence in these new financial intermediaries.  If we can’t, then reports of a revolution would seem greatly exaggerated.

Dan Awrey is an Associate Professor of Law and Finance and Academic Director of the MSc in Law and Finance programme.



For more on Smart Contracts, see Smart Contracts: Bridging the Gap Between Expectation and Reality - also available on the Oxford Business Law Blog.


With the support of