Look Out for Each Other. That's Why We Built Crew Mode.
When friends go out together, looking out for one another should be part of the plan.
Notice when someone seems unusually disoriented. Check in when a friend disappears from the group. Take them seriously when they say something feels wrong. Make sure everyone has a way home, and confirm they arrived.
Friends who pay attention and act can be one of the strongest sources of protection on a night out. They know each other, notice changes, and can get help when someone cannot ask for it themselves.
That is why we created Crew Mode.
Recent reporting about sexual assault allegations at Cornell has brought renewed attention to campus safety, institutional accountability, and the role of alcohol and drugs in sexual violence. These are serious allegations involving real people. They should not be reduced to a product pitch.
They are also a reason to talk plainly about how we care for one another, what technology can help with, and what it cannot do.
Make looking out for each other part of going out
Before heading out, decide how you will stay in touch and get home. During the night, keep track of your friends and check in when plans change.
- Stay connected. Tell the group when you leave or change locations.
- Check on changes in behavior. Sudden confusion, extreme drowsiness, or difficulty standing deserves attention, whatever the cause.
- Stay with a friend who needs help. Do not leave them alone or let them leave with someone they do not trust.
- Get medical help promptly. If someone is unresponsive, has trouble breathing, or appears seriously ill, call 911 in the US. Do not assume they can sleep it off.
- Confirm arrival. A ride being booked is not the same as a friend getting home.
None of these steps guarantees safety. Being harmed is never evidence that someone failed to take the right precautions. Responsibility for assault belongs to the person who commits it, not the victim or their friends.
Looking out for each other means offering support, not assigning blame.
Why we created Crew Mode
A group chat can help friends make plans. Once everyone is out, keeping up with changing locations and checking on one another can get harder.
Crew Mode brings group planning, chat, and location-sharing tools together to help friends coordinate. Members control their routine location sharing, and the app includes a Flare action for alerting their Crew when they need help.
The purpose is to support the care friends already give each other.
A location pin cannot tell you whether someone is okay. A message cannot replace a call when you are worried. An app is not an emergency service, and a Flare does not contact emergency dispatch.
Check-ins are manual. Use the tools to stay connected, then follow through. Check in. Call. Find your friend. Get help.
What the research tells us about spiking
There is no single figure here that tells you the chance of being drugged on a night out. Surveys measure what people say happened to them. Police statistics measure what was reported. Neither captures every incident, and neither should be read as a confirmed count of covert drugging.
7.8% in a three-university survey
A study of 6,064 students at three US universities reported that 462 students said they had been drugged. The researchers reported a prevalence of 7.8%, with 539 incidents described by those respondents. The study was published online in May 2016 and appeared in the April 2017 issue of Psychology of Violence.
What that means: self-reported experiences in that study, not laboratory-confirmed drugging, a current national estimate, an annual rate, or the risk of one night out. The authors explicitly said they could not establish whether respondents had actually been drugged.
6,732 police reports in a UK reporting year
The UK Home Office reported that police received 6,732 reports of spiking in the year ending April 2023. Of those reports, 957 related to needle spiking.
What that means: a count of reports covering several forms of spiking, not 6,732 confirmed drink-spiking cases. It is not a US figure or a population prevalence rate, and this single reporting year does not establish a trend.
These findings give the concern substance without giving us a precise personal risk estimate. They also do not measure how much going out with friends reduces risk. We recommend looking out for each other as a practical way to notice trouble and respond, not as a statistically proven guarantee.
Whatever the cause, someone who suddenly becomes seriously unwell needs help. Friends do not need to establish that a drink was spiked before acting.
We also need better ways to study suspected spiking
Friends can respond to what they see in the moment. Researchers face a different problem: understanding patterns in records that may be incomplete or ambiguous.
A report of suspected drink spiking is not the same as a toxicologically confirmed case. A medical call involving an unconscious person does not establish why they became unconscious. Coerced drug use and covert drink spiking are also different circumstances, even when both raise serious concerns about consent and harm.
Those distinctions matter when discussing the Cornell allegations and when analyzing public data.
To make one research approach available for inspection, we published the Spiking Detection Toolkit, an open-source project under the MIT license.
The toolkit includes code and queries for screening available public records for descriptions and indicators that may warrant further review. It separates stronger narrative signals from weaker contextual ones and includes methods for handling uncertainty, activity levels, small samples, and ambiguous locations.
Open code does not mean proven results
The toolkit is a research implementation, not a validated test for whether someone was drugged.
Its default screening weights are starting assumptions. They need to be calibrated against locally reviewed records before results can be trusted. Tests using simulated data help assess the statistical implementation, but they do not establish real-world detection accuracy.
We have not analyzed the Cornell case with this toolkit. We are not claiming it could have predicted or prevented what is alleged to have happened.
We also do not present its exploratory outputs as confirmed incident totals, personal risk estimates, or reliable rankings of individual venues.
Making the work open lets others inspect the rules, question the assumptions, reproduce the methods, and identify mistakes. We welcome that scrutiny. Researchers, developers, and public-health practitioners can review the methodology and contribute through the repository.
The research matters. So does the friend beside you.
Better data can support research and inform prevention efforts. It cannot watch out for your friend tonight.
Go together. Check in. Take concerns seriously. Make sure everyone gets home.
We built Crew Mode to help friends put that care into practice.
Android beta access requires approval. For practical guidance and confidential support resources, read our drink-spiking and safety guide.
Sources
- Swan and colleagues, Just a Dare or Unaware? Psychology of Violence, online 2016; journal issue 2017. APA summary.
- UK Home Office, Spiking: factsheet, December 2023.
- NPR, Cornell agrees to an independent probe into its handling of a fraternity rape case, updated September 30, 2026.
- Cornell University, Resources and support for sexual assault and misconduct, September 29, 2026.
- Spiking Detection Toolkit, README, methodology, and MIT license.
- RAINN, drug-facilitated sexual assault, response and care guidance.